Mitochondrial Metabolism in Developing Embryos of Brassica napus

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 281, NO. 45, pp. 34040 –34047, November 10, 2006 © 2006 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

Mitochondrial Metabolism in Developing Embryos of Brassica napus*□ S⽧

Received for publication, June 30, 2006, and in revised form, September 11, 2006 Published, JBC Papers in Press, September 12, 2006, DOI 10.1074/jbc.M606266200

Jo¨rg Schwender‡1, Yair Shachar-Hill§, and John B. Ohlrogge§ From the ‡Biology Department, Brookhaven National Laboratory, Upton, New York 11973 and the §Plant Biology Department, Michigan State University, East Lansing, Michigan 48824

Directly or indirectly plant seeds provide most of the food consumed by humans. Although the metabolism of developing seeds has been extensively studied, quantitative understanding of fluxes through central metabolism is still quite limited. Brassica napus (canola, oilseed rape) is a major oil crop and is ame-

* This

work was supported by National Science Foundation Grant MCB 0224655, United States Department of Agriculture Grant 2003-3532112935, and by a Laboratory Directed Research and Development Award at the Brookhaven National Laboratory under contract with the United States Department of Energy (to J. S.). Acknowledgment is also made to the Michigan Agricultural Experiment Station for its support of this research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ⽧ This article was selected as a Paper of the Week. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental text, Figs. S1–S6, Tables S1–S10, and references. 1 To whom correspondence should be addressed: Biology Dept., Brookhaven National Laboratory, Bldg. 463, Upton, NY 11973. Tel.: 631-344-3797; Fax: 631-344-3407; E-mail: [email protected].

34040 JOURNAL OF BIOLOGICAL CHEMISTRY

nable to detailed quantitative flux analysis under conditions that closely mimic in planta seed development (1). The main storage compounds in seeds of B. napus are oil (triacylglycerols) and proteins, which are synthesized by the developing embryo from sugars and amino acids taken up from the surrounding endosperm liquid. Developing seeds of B. napus have also been the subject of numerous biochemical studies and are a model for oil accumulating seeds (2–13). In previous studies we have used intact developing embryos to make a quantitative analysis of steadystate metabolic fluxes during the conversion of carbohydrates to fatty acids in vivo (1, 14 –17). These studies introduced a labeling approach using multiple carbon sources (1), quantified the contribution of the oxidative pentose phosphate pathway to biosynthetic NADPH demands (14), and revealed that ribulose1,5-bisphosphate carboxylase/oxygenase (Rubisco)2 operates in a novel context of carbohydrate conversion to fatty acids, bypassing glycolytic reactions and increasing the efficiency of seed carbon metabolism (16). However, a detailed description of fluxes through mitochondrial metabolism in B. napus or other developing seeds is lacking. Several functions of mitochondrial metabolism known in plants could be quantitatively relevant for seed development in B. napus. In root tips and cell suspension cultures substantial flux around the mitochondrial tricarboxylic acid cycle results in more than 20% of all the glucose carbon entering catabolism being oxidized to CO2 (calculated from data in Refs. 18 and 19). This suggests that most of the cellular ATP is generated by oxidative phosphorylation driven by the oxidation of acetylCoA in the tricarboxylic acid cycle. In developing seeds this metabolism could likewise provide much of the ATP that is needed in substantial quantities for the synthesis of storage proteins and oil. In addition, the cytosolic elongation of oleic acid (C18:1) to C20:1 and C22:1 fatty acids in developing seeds requires mitochondrial citrate as a precursor (1, 20, 21). This demand could be large since in many Brassica species erucic acid (C22:1) is the most abundant fatty acid stored in the seed. Furthermore, in leaves and roots mitochondrial conversion of OAA and Pyr to KG is essential to primary nitrogen assimilation. Because developing embryos receive glutamine and other 2

The abbreviations used are: Rubisco, ribulose-1,5-bisphosphate carboxylase/oxygenase; ICDH, isocitrate dehydrogenase; KG, ␣-ketoglutarate; OAA, oxaloacetate; PEP, phosphoenolpyruvate; Pyr, pyruvate; TBS, N,Otert-butyldimethylsilyl; GC-MS, gas chromatography-mass spectrometry; FW, fresh weight. Suffixes: “cyt,” “pl,” and “mit” designate the cytosolic, plastidic, and mitochondrial subcellular compartments, respectively (e.g. Alacyt for cytosolic alanine).

VOLUME 281 • NUMBER 45 • NOVEMBER 10, 2006

Downloaded from http://www.jbc.org/ by guest on May 16, 2016

The metabolism of developing plant seeds is directed toward transforming primary assimilatory products (sugars and amino acids) into seed storage compounds. To understand the role of mitochondria in this metabolism, metabolic fluxes were determined in developing embryos of Brassica napus. After labeling with [1,2-13C2]glucose ⴙ [U-13C6]glucose, [U-13C3]alanine, [U-13C5]glutamine, [15N]alanine, (amino)-[15N]glutamine, or (amide)-[15N]glutamine, the resulting labeling patterns in protein amino acids and in fatty acids were analyzed by gas chromatography-mass spectrometry. Fluxes through mitochondrial metabolism were quantified using a steady state flux model. Labeling information from experiments using different labeled substrates was essential for model validation and reliable flux estimation. The resulting flux map shows that mitochondrial metabolism in these developing seeds is very different from that in either heterotrophic or autotrophic plant tissues or in most other organisms: (i) flux around the tricarboxylic acid cycle is absent and the small fluxes through oxidative reactions in the mitochondrion can generate (via oxidative phosphorylation) at most 22% of the ATP needed for biosynthesis; (ii) isocitrate dehydrogenase is reversible in vivo; (iii) about 40% of mitochondrial pyruvate is produced by malic enzyme rather than being imported from the cytosol; (iv) mitochondrial flux is largely devoted to providing precursors for cytosolic fatty acid elongation; and (v) the uptake of amino acids rather than anaplerosis via PEP carboxylase determines carbon flow into storage proteins.

Mitochondrial Metabolism in B. napus Embryos amino acids from the mother plant, it is unclear how their mitochondria are involved in amino acid metabolism. Finally, in leaves mitochondria carry large fluxes through the photorespiratory pathway (22). However, photorespiration does not occur in B. napus embryos (1) probably because of high CO2 and low oxygen concentrations (17, 23). Building on our earlier studies on central carbon metabolism in B. napus embryos (1, 14, 16), this study focuses on the quantification of fluxes through mitochondrial metabolism and their integration with the major fluxes of de novo fatty acid synthesis in plastids and fatty acid elongation in the cytosol. The resulting map of mitochondrial fluxes shows that mitochondrial metabolism in developing B. napus embryos differs fundamentally from that described for other eukaryotic cells.

EXPERIMENTAL PROCEDURES

NOVEMBER 10, 2006 • VOLUME 281 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY

34041

Downloaded from http://www.jbc.org/ by guest on May 16, 2016

Chemicals—D-[1,2-13C2]Glucose, D-[U-13C6]glucose, L-[UC5]Gln, L-(amide)-[15N]Gln and L-(amino)-[15N]Gln (all 99% 13 C or 15N abundance) were purchased from Isotec (Miamisburg, OH). L-[U-13C3]Ala and L-15N-Ala were from Cambridge Isotope Laboratories (Andover, MA). Embryo Culture—Oilseed rape plants (B. napus L., cv. Reston) were grown as described previously (1, 14). Siliques were harvested 20 days after flowering, and embryos at the early stage of oil accumulation (0.5–1 mg of FW) were immediately dissected under aseptic conditions and transferred into a liquid culture medium with inorganic nutrients (1), 20% polyethylene glycol 4000, and carbon and nitrogen sources sucrose (80 mM), Glc (40 mM), Gln (35 mM), and Ala (10 mM). Embryos were grown for 14 days in 5 ml of growth medium in 250-ml Erlenmeyer flasks closed with cotton plugs under low light conditions (21 °C, continuous light of 50 ␮mol m⫺2 s⫺1). Embryos of 5 mg of FW doubled in weight in 50 ⫾ 11 h (n ⫽ 5). For 13C labeling, isotopes were selected to increase the sensitivity of flux determinations as described by Schwender et al. (15) following principles outlined by Mollney et al. (24). Glucose, Ala, or Gln were replaced with [1,2-13C2]glucose/[U-13C6]glucose (1:1 mol/mol ratio), [U-13C3]Ala, or [U-13C5]Gln (1:1 (mol/mol) mixture with unlabeled Gln), respectively. In addition to 13C labeling, experiments using (amide)-[15N]Gln, (amine)[15N]Gln, or [15N]Ala were performed to assess the contribution of Gln and Ala to protein synthesis independently from the 13 C experimental results as well to assess the exchange of 15N label between proteinogenic amino acids by transaminase activity (see supplemental text (section 4.2)). Extraction of Lipids and Proteins—Embryos were collected, frozen immediately, and kept at ⫺20 °C until extraction. Embryos were ground in a glass tissue grinder at 4 °C and proteins were extracted in a buffer containing sodium phosphate, pH 7.5 (10 mM), and NaCl (500 mM, 1). Storage lipids were extracted by adding hexane/diethylether (1:1, v/v) during protein extraction. Extracted proteins were precipitated by the addition of 0.10 volume of 50% trichloroacetic acid. Derivatization of Lipids and Proteinogenic Amino Acids—After extraction, lipids were reduced under hydrogen (platinumIV-oxide), and the saturated fatty acids were transesterified to form methyl esters as described previously (14). Proteins were 13

hydrolyzed in 6 N HCl, and the amino acids were derivatized to N,O-tert-butyldimethylsilyl (TBS) derivatives as described previously (14). Measurement of Labeling in Lipids and Amino Acids by GC-MS—Fatty acid methyl esters and TBS amino acids were analyzed with a HP 5890 II (Hewlett-Packard) gas chromatograph-mass spectrometer (HP 5972 quadrupole MS). The carrier gas was helium at a flow rate of 1 ml/min. DB23 and DB1 columns (30 m ⫻ 0.25 mm; J&W Scientific, Folsom, CA) were used for analyzing fatty acid methyl esters and TBS amino acids, respectively. The relative abundance of mass isotopomers in selected fragments of each analyzed derivative was measured using selected ion monitoring (14). Correction for the occurrence of 13C in derivative parts of the molecules and for heavy isotopes in heteroatoms (carbon, hydrogen, oxygen, nitrogen, silicon) at their natural abundances was performed as described earlier (14). MS measurements from three injections and analyses from each sample were averaged. For saturated fatty acid methyl esters the McLafferty fragment (m/z 74) was analyzed to determine the fractional labeling in the fragment comprising carbons one and two of each fatty acid (14). Glycerol remaining from the transesterification of lipids was analyzed as the trifluoroacetic acid ester (14). Glucose was released from starch by hydrolysis and analyzed as Glc methoxime penta-acetate (14). For each GC-MS chromatogram of amino acid derivatives 136 mass isotopomer fractions were recorded, whereas for fatty acid methyl esters, glycerol, and glucose derivatives 29 mass isotopomer fractions were monitored. Computer-assisted Flux Estimation Using Labeling and Biomass Constraints—Flux analysis of central metabolism was based on the labeling signatures measured in metabolic end products, considering that the carbon atoms incorporated into amino acids and fatty acids can be traced back to the structures of a number of central intermediates (see supplemental text (section 1.2)). A biochemical reaction network of central carbon metabolism was derived from the literature that describes enzymes of central metabolism in B. napus embryos (see supplemental text (section 1.3)) as well as from the biochemical pathway data base for Arabidopsis (AraCyc). Implementation of the isotopomer model, isotopomer balancing, flux parameter fitting, and statistical analysis were performed using the software package 13CFLUX obtained from Dr. W. Wiechert (Department of Simulation, University of Siegen, Germany; see Ref. 25 and referencess cited therein). In applying the modeling software the flux values in central metabolism were constrained by: 1) the topology of the metabolic network and its stoichiometric relations at metabolic steady state; 2) assuming irreversibility of particular reactions according to the thermodynamics of the reaction; 3) the biomass composition of the embryos, which defines the net fluxes of metabolites into amino acid and fatty acid synthesis (mass balancing); 4) the results of 15N labeling experiments, which quantified the reversibilities of Ala and Glu aminotranferases; and 5) the measured 13C labeling patterns in metabolites from independent labeling experiments using [1,2-13C2]glucose/[U-13C6]glucose, [U-13C5]Gln and [U-13C3]Ala, respectively. By varying the values for the 17 freely variable fluxes in the system a best fit between predicted and measured mass isotopomer abundances was obtained by mini-

Mitochondrial Metabolism in B. napus Embryos

34042 JOURNAL OF BIOLOGICAL CHEMISTRY

soybean embryos. The labeling of the product succinate via the two alternative routes would be identical and they cannot be distinguished based on the labeling data of this study or of Sriram et al. (27).

RESULTS Experimental Strategy—Previous studies of in vivo flux in plant central carbon metabolism using steady-state stable isotope labeling have been based almost exclusively on [13C]glucose or [13C]sucrose feeding, generally using a single labeled substrate. Our previous studies showed that because of the complexities introduced by the subcellular compartmentation of metabolism, experiments using only [13C]glucose or [13C]sucrose have limited ability to determine several key mitochondrial fluxes. Therefore, to obtain reliable flux measurements, embryos were labeled in separate experiments with [U-13C3]alanine, [U-13C5]glutamine, [15N]alanine, (amino)[15N]glutamine, or (amide)-[15N]glutamine in addition to [1,213 C2]glucose ⫹ [U-13C6]glucose. After each labeling experiment, protein and oil were extracted, hydrolyzed, and analyzed by GC-MS to determine fractional 13C or 15N enrichments for selected fragment ions. With labeling patterns obtained from the different 13C label experiments, best-fit values for the metabolic fluxes were derived using computer-assisted modeling (see “Experimental Procedures” and supplemental text (section 2)). The strategy of combining data from experiments with different labeled precursors increased the reliability and accuracy of the flux determinations. Unbiased flux parameter fitting consistently converged to one best fit only if the labeling data derived from the three 13C labeling experiments (using [1,2-13C2]glucose ⫹ [U-13C6]glucose, [U-13C3]alanine, or [U-13C5]glutamine) and 15N labeling experiments were combined. For the reactions of mitochondrial metabolism statistical accuracy was increased by the combined data. For example, if only the data from [13C]glucose labeling is considered, the standard deviations for the net fluxes vME, vICDH, vGAT, vPyr2, and vPEPC (Table 1) were determined to be 57, 454, 29, 27, and 38% of the flux value, respectively (see supplemental text (section 4.3)). However, with the combined labeling data these S.D. values were reduced to 14%, 216, 12, 9, and 10%, respectively. Overall 30 out of 31 net fluxes were determined with a confidence of better than ⫾ 20% of their value. The reversibility of the isocitrate dehydrogenase reaction (see below) was unambiguously identified only after labeling with [U-13C5]Gln. In addition, and importantly for model validation, the interconversion of amino acids and their corresponding ␣-keto acids (e.g. Pyr/Ala, KG/Glu) was assessed by 15N labeling experiments. Flux parameter fitting consistently converged to unique solutions for most free net fluxes. However, for the import of pyruvate into the plastid (vPyr1) more detailed inspection of independent sets of optimized flux values revealed that the isotopomer model explains the MS data equally well with different combinations of the reversible interconversion of Alac and pyruvatec (vAATXCH) and vPyr1 (see Fig. 1) and that these fluxes cannot be resolved by 13 C data alone. Therefore, since vAATXCH exchanges amino VOLUME 281 • NUMBER 45 • NOVEMBER 10, 2006

Downloaded from http://www.jbc.org/ by guest on May 16, 2016

mizing the sum of squared differences. The reliability of the flux estimates was tested by starting the flux parameter fitting process 100 times independently with starting values for the free flux parameters randomly chosen within the feasible flux space. Further details of the flux estimations are described under “Results” and in the supplemental text (section 4). Statistical Analysis—Using the 13CFlux statistical software tool, the S.D. values for the flux values (see Table 1) were derived from the S.D. values of the biomass and MS measurements. As the MS data from three independent 13C labeling experiments ([1,2-13C2]glucose/[U-13C6]glucose, [U-13C5]Gln, and [U-13C3]Ala) were combined in the model, the biological and analytical variability resulting from triplication of the labeling experiment are present in the data set. However, since the three different labeling patterns cannot be directly averaged, values for the S.D. in the MS measurements were estimated as detailed in the supplemental text (section 5.2), and S.D. in the fluxes calculated on this basis (Table 1). Based on the range of values for the S.D. observed in the MS measurement data in comparable labeling experiments the ␹2 value is ⬎35 (see supplemental text (section 5.2)). The ␹2 test for goodness of fit is expected to pass since 131 is the maximal allowed value for ␹2 on a 90% significance level. In addition, the robustness of the system was tested by removal of 13 of the 164 MS measurements from the data set, causing 50% of the ␹2 error. With the reduced data set flux parameter fitting still consistently converged to the same solution (see supplemental text (section 4.2)). This demonstrates the robustness of the model achieved by overdetermination. Therefore we judge the model predictions (flux values, Table 1) to reasonably explain the experimental data. Design and Validation of the Flux Model—The following reactions were tested in the model to determine whether they improved the agreement between measured and model-predicted labeling levels or if they led to uncertainties in the routes of metabolic flux: 1) isocitrate lyase together with malate synthase, defining a glyoxylate bypass; 2) malic enzyme as a reversible reaction; 3) export of plastidic AcCoA into the cytosol. In each case flux parameter fitting consistently assigned very small fluxes to the added reactions and the quality of the fit was not increased by their inclusion (data not shown). Similarly, flux through reactions of gluconeogenesis (conversion of OAA to PEP) could be excluded based on labeling in metabolites made from OAA and PEP. The transformation of glycolate to serine, which carries a very high flux in leaf mitochondria during light respiration (22), is not a significant flux in developing B. napus embryos (1). As outlined in the results section, the reversibility of isocitrate dehydrogenase (ICDH) was included based on fitting and further experiments. Consideration of Alternative Compartmentation—In some cases there are alternative pathways that would produce identical labeling signatures in end metabolites. In the flux model ICDH was assumed to be mitochondrial, although cytosolic and plastidic isoenzymes exist in plants (26). Similarly, the conversion of KGmit to Succmit is assumed to be performed by mitochondrial KG-DH. However this conversion could also take place via the GABA shunt, as assumed by Sriram et al. (27) for

Mitochondrial Metabolism in B. napus Embryos TABLE 1 Values of net and exchange fluxes (ⴞS.D.) for B. napus embryos as derived by flux parameter fitting using the fractional enrichment in amino acids, carbohydrate, and fatty acids of 13C and 15N labeling experiments The fluxes are based on the observed growth rate of ␮ ⫽ 0.014 h⫺1 at 5 mg of embryo FW. For exchange fluxes unsymmetrical 68% confidence intervals are given in parentheses instead of S.D. 3-PGA, 3-phosphoglycerate. Flux name

Reaction

Fluxes

Enzyme name(s)/comment

Net flux

Exchange flux

nmol h⫺1 mg FW⫺1

Glc uptake Ala uptake Gln uptake 3-PGA 3 PEP

vPKc vPKpl vPyr1 vPyr2 vPDHpl vAAT vGlnGlu vGAT

PEP 3 Pyrcyt PEP 3 Pyrpl Pyrcyt 3 Pyrpl Pyrcyt 3 Pyrmit Pyrpl 3 AcCoApl Alacyt 3 Pyrcyt Glncyt 3 Glucyt Glu 3 KGmit

vPEPC vMEmit vCS vPDHmit vACL vICDH vKDH

PEP ⫹ CO2 3 OAA/malate Malate 3 Pyrmit ⫹ CO2 OAA ⫹ AcCoAmit 3 Cit Pyrmit 3 AcCoAmit ⫹ CO2 Citcyt 3 AcCoAcyt ⫹OAA Citmit 3 KG ⫹ CO2 KG 3 Fum ⫹ CO2

vFM

Fum 3 OAA/Mal

58.3 ⫾ 4.2 7.4 ⫾ 0.7 4.3 ⫾ 0.3 83 ⫾ 8.3

Uptake of glucose and sucrose from medium Uptake of Ala from medium Uptake of Gln from medium Phosphoglyceromutase, enolase (plastidic plus cytosolic fluxes combined) Cytosolic pyruvate kinase Plastidic pyruvate kinase Transport of pyruvate into plastids Transport of pyruvate into mitochondria Plastidic pyruvate dehydrogenase complex Ala:KG aminotransferase Glutamate synthase Transaminases, Glutamate dehydrogenase, dicarboxylate transporter PEP carboxylase

17.9 ⫾ 2.0 60.0 ⫾ 6.5 21.0 ⫾ 2.6 3.5 ⫾ 0.3 77.0 ⫾ 8.9 6.6 ⫾ 0.7 3.0 ⫾ 0.2 1.4 ⫾ 0.2 3.6 ⫾ 0.3 2.3 ⫾ 0.3 5.8 ⫾ 0.6 5.8 ⫾ 0.6 5.9 ⫾ 0.6 ⫺0.1 ⫾ 0.2 1.3 ⫾ 0.2

Mitochondrial malic enzyme Citrate synthase Mitochondrial pyruvate dehydrogenase complex Cytosolic ATP:Citrate lyase Aconitase, isocitrate dehydrogenase Ketoglutarate dehydrogenase Succinyl-CoA synthetase Succinate dehydrogenase Fumarase, malate dehydrogenase

1.3 ⫾ 0.2

Fluxes derived from biomass proportions (biomass constraints) vHP_St Hexose phosphates into starch and cell wall polymers vFAS AcCoA 3 FA Total synthesis rate of fatty acids Plastidic fatty acid synthesis to chain length of 16 and AcCoApl 3 FA vFASpl 18 carbons vFASc Flux of AcCoA into cytosolic elongation AcCoAcyt ⫹ C18 3 C20 of C18 fatty acids to C20 and C22 2 AcCoAcyt ⫹ C18 3 C22 vAla_P Ala into protein Protein bound Ala, assumed to be of cytosolic origin Pyrpl into protein Plastidic branched chain amino acid synthesis (Val, vPyrpl_P Leu, Ile, Lys) vPEP_P PEP into protein Plastidic aromatic amino acid synthesis (Phe, Tyr, Trp) vAcCoApl_P AcCoApl into protein Leu biosynthesis vAsp_P OAA into protein Asp and derived amino acids: Asn, Thr, Met, Ile, Lys vGlu_P Glu into protein Includes also Pro and Arg, derived from Glu vGln_P Gln into protein a

190a (106, 274) 1.3 (0.7, 1.8) 4.7 (3.6, 5.8)

Downloaded from http://www.jbc.org/ by guest on May 16, 2016

vuptGlc vuptAla vuptGln vPGM

4.5 (3.5, 5.4)

30 (14, 47)

15.9 ⫾ 1.4 82.0 ⫾ 8.9 76.1 ⫾ 8.9 5.9 ⫾ 0.6 0.7 ⫾ 0.1 4.0 ⫾ 0.6 1.6 ⫾ 0.2 0.9 ⫾ 0.1 2.6 ⫾ 0.3 1.6 ⫾ 0.2 1.3 ⫾ 0.2

Values were determined by 关15N兴Ala labeling experiments (see supplemental text (section 4.2)).

tein) were derived by applying isotopic steady state and the stoichiometry shown in Fig. 1. vAATXCH (%[15N]Alamedium)⫺(%[15N]Alaprotein) ⫽ vuptAla (%[15N]Alaprotein)⫺(%[15N]Proprotein)

FIGURE 1. Interconversion of cytosolic Ala and pyruvate by alanine aminotransferase. By using 15N-labeled Ala and unlabeled Gln in embryo cultures the 15N enrichment in protein-bound Ala (Alaprotein) is dependent on the ratio of the uptake of [15N]Ala (vuptAla) and the conversion of pyruvate to Ala (vAATXCH). Given 15N enrichment measurements in Ala in the medium, Alaprotein and Proprotein (see supplemental text (section 4.2)) the exchange flux vAATXCH can be determined relative to the rate of alanine uptake (see “Results”).

groups by transamination, it was assessed based on the results of labeling with 15N-Ala and 15N-Gln. The steady state equations for the ␣-nitrogen of cytosolic Ala (in proNOVEMBER 10, 2006 • VOLUME 281 • NUMBER 45

(Eq. 1)

Using Equation 1 and the 15N measurements shown in supplemental Table S7a, the value for vAATXCH was determined as 25.8 ⫾ 11.4 times the input flux of alanine (vuptAla), which corresponds to essentially complete isotopic equilibration between cytosolic pyruvate and alanine. In the 13C isotopomer model the exchange flux vAATXCH was then fixed to be 25.8 times vuptAla. After this determination of vAATXCH the optimizer algorithm was started 100 times with randomly assigned starting values for the free fluxes, and the optimization algorithm consistently converged to one solution with the lowest sum of weighted squared deviations (residuum) for the free net fluxes, suggesting that this solution is a global optimum. This approach allowed vPyr1 to be well determined as being 21.0 ⫾ 2.6 nmol h⫺1 mg FW⫺1 (Table 1). JOURNAL OF BIOLOGICAL CHEMISTRY

34043

Mitochondrial Metabolism in B. napus Embryos

34044 JOURNAL OF BIOLOGICAL CHEMISTRY

VOLUME 281 • NUMBER 45 • NOVEMBER 10, 2006

Downloaded from http://www.jbc.org/ by guest on May 16, 2016

Of critical importance for the validity of the tricarboxylic acid cycle fluxes is that label measured in protein-derived Asp represents the label of the mitochondrial OAA pool (Fig. 2). As detailed in the supplemental text (section 6.4) this assumption is justified as extensive symmetric randomization of label was observed in protein-derived Asp, making evident an intensive interconversion between cytosolic OAA and mitochondrial fumarate via mitochondrial transporters, malate dehydrogenase and fumarase. This symmetric randomization has also been observed before in other plant systems (18, 28, 29). In addition, the flux parameter fitting consistently assigned a high exchange flux to the interconver- FIGURE 2. Net fluxes in the central metabolism network of developing B. napus embryos. The arrow sion of fumarate and OAA (Table 1), thicknesses are proportional to net fluxes of carbon. Open arrowheads indicate reactions modeled as reversible. values of fluxes see Table 1. Only fluxes downstream of 3-phosphoglycerate are shown. Fluxes to amino isotopically equilibrating both For acids and triacylglycerols designate deposition of amino acids and fatty acids as storage protein and seed oil, pools. respectively. Metabolites for which the isotopomer composition was determined by GC-MS and used for flux Fluxes through Mitochondrial parameter fitting are underlined. Dashed boxes unify metabolites across compartments, which are not resolved by the model (see “Results”). Cit, citrate; 3PGA, 3-phosphoglycerate; FA; fatty acid; Fum, fumarate; Mal, malate. Metabolism—The current study focuses on mitochondrial metaboSeveral features of mitochondrial metabolic flux shown in lism as it relates to the production of precursors for the major storage products, oil and protein, and, via oxidative phospho- Fig. 2 are either novel or highly unusual. These are discussed rylation, to ATP production. The flux values obtained are below. shown in Table 1 and Fig. 2. Several reactions shown in Fig. 2 DISCUSSION describe two or more sequential enzyme reactions that are Unconventional Mitochondrial Metabolism in Developing B. considered together (see Table 1). Although the reactions of glycolysis, the pentose phosphate pathway, and Rubisco napus Embryos—Studies in plants have assigned mitochondrial were included in the full analysis, Fig. 2 and the discussion metabolism central roles in respiration, photorespiration and focus on the reactions and metabolic pools downstream of biosynthesis as well as in gluconeogenesis, the glyoxylate cycle, and amino acid degradation (22, 30, 31). Several canonical flux phosphoglycerate. Isocitrate Dehydrogenase Is Reversible in Vivo—The ICDH patterns are recognized: (a) primarily respiratory in heterotroreaction was initially assumed to be irreversible. However, phic tissues; (b) primarily photorespiratory in leaves and other autotrophic tissues of C3 plants; and (c) primarily gluconeoduring flux parameter fitting, the labeling observed in Pro genic in germinating oilseeds, using the glyoxylate bypass. Met(whose labeling represents that in ␣KG) was much better abolic flux analysis in heterotrophic plant tissues using stable explained by the model if ICDH was allowed to be reversible isotope labeling has confirmed the first of these functions (18, (fitting results not shown). To test this possibility, embryos 19, 27–29, 32). Our findings (Table 1 and Fig. 2) show that were grown with [U-13C5]Gln (50% 13C enrichment), and mitochondrial metabolism of developing B. napus embryos diffree organic acids were extracted and analyzed as TBS deriv- fers substantially from that previously described in other plant atives. In the mass spectrum of citrate the molecular ion systems or indeed animal or microbial cells. The following secincluded an m⫹5 peak of ⬎10% abundance (Fig. 3). Based on tions discuss these novel features. the mass isotopomer distribution in Asp (representing Flux around the Tricarboxylic Acid Cycle Is Absent—In OAA) and in the C1 ⫹ C2 fragment of C22 fatty acids developing B. napus embryos there is no cyclic flux through the (derived from mitochondrial acetyl-CoA), if citrate was reactions of the tricarboxylic acid cycle (Fig. 2). Although there formed from OAA and Ac-CoA, the m5 peak in citrate(1– 6) is a significant forward flux through the other reactions, there is should be ⬍2% (see Fig. 3). Therefore the intensity of this almost no net flux or possibly a small net reverse flux through peak cannot be explained by the action of citrate synthase. the ICDH reaction. Furthermore, only 3.7% of the amount of Instead, the conversion of [U-13C5]KG (the labeling that was hexose carbon that is catabolized is released as CO2 by B. napus determined from the mass isotopomer distribution of Pro) to embryo mitochondrial decarboxylase activities (pyruvate dehycitrate by reverse flow through ICDH accounts for the detec- drogenase, malic enzyme, ICDH, and ketoglutarate dehydrogenase combined). This is in striking contrast to other plant tion of this citrate isotopomer containing five 13C atoms.

Mitochondrial Metabolism in B. napus Embryos

systems where 19 –27% of hexose that enters catabolism is oxidized to CO2 by these reactions (see Table 2). Mitochondrial Substrate Oxidation Contributes Little to ATP Production in B. napus Embryos—Mitochondrial substrate oxidation and biosynthetic ATP demands were calculated from the flux values in Table 1 (supplemental Table S9). For the mitochondrial oxidative phosphorylation all mitochondrial NADH- and FADH-producing reactions were balanced assuming ATP production stoichiometries of 3 for NADH and 2 for FADH. The biosynthetic ATP demands were estimated as follows. During fatty acid synthesis acetyl-CoA carboxylase requires one ATP per acetyl unit incorporated into fatty acids. Protein synthesis needs ATP for the polymerization of amino acids. The cost for protein synthesis was assumed to be 4.3 mol of ATP per amino acid incorporated (based on estimates for protein biosythesis in microorganisms; Ref. 33). This does not include ATP demands for amino acid biosynthesis. Based on these assumptions ATP produced by mitochondrial substrate oxidation and oxidative phosphorylation is 35 nmol h⫺1 mg FW⫺1, while the demand is 159 nmol h⫺1 mg FW⫺1. Since maximal possible phosphate/oxygen ratios were assumed and the biosynthetic ATP demand reflects minimum requirements, the resulting ratio of mitochondrial ATP production to biosynthetic ATP demand of 0.22 is likely to be an overestimate. Therefore we conclude that mitochondrial substrate oxidation with subsequent oxidative phosphorylation NOVEMBER 10, 2006 • VOLUME 281 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY

34045

Downloaded from http://www.jbc.org/ by guest on May 16, 2016

FIGURE 3. Label in citrate reveals reverse in vivo flux of isocitrate dehydrogenase. After labeling of embryos with [U-13C5]glutamine (50% 13C5), mass isotopomer fractions were measured for the carbon skeletons of citrate, Asp, Pro, and C22(1–2) (gray bars), which represent labeling in citrate, OAA, KG, and cytosolic Ac-CoA, respectively. Black bars represent mass isotopomers expected in citrate if it was formed only from oxaloacetate and Ac-CoA (as predicted by probabilistic equations using the mass isotopomer fractions in Asp(1– 4) and C22(1–2)). Fragments analyzed by GC-MS of t-butyldimethylsilyl (TBS) derivatives of organic acids and methylesters of fatty acids were: Citrate (1– 6), [citrate-TBS4]-57, m/z 459.2; Asp(1– 4), [Asp-TBS3]-57, m/z 418.2; Pro(1–5): [Pro-TBS3]-57, m/z 286.17 (data not shown); C22(1–2), [C22: O-methylester]-180, m/z 74.1.

produces at most 22% of the ATP required for the biosynthetic needs of developing B. napus embryos. Thus mitochondrial substrate oxidation makes only a small contribution to the energy required for storage product accumulation and other cellular processes in B. napus embryos. The main amount of cellular ATP has to be provided by photosynthetic light reactions (12, 17), mitochondrial oxidative phosphorylation using cytosolic NADH, or substrate-level phosphorylation in glycolysis. Isocitrate Dehydrogenase Is Reversible in Vivo—The data of Fig. 3 demonstrate the in vivo reversibility of ICDH in developing B. napus embryos, an aspect of tricarboxylic acid metabolism that has not been reported previously for plants. As a consequence of this reversibility there are more options for directing flux through mitochondrial metabolism since reversibility allows net reverse flux (Table 1 and Fig. 2). Reversibility of the ICDH reaction has been reported previously for mammalian liver (34), and it has been shown that failure to consider reversibility for the ICDH reaction can lead to incorrect quantification of mitochondrial fluxes (35). The in vivo reversibility of ICDH in B. napus may in part be attributed to the very high tissue concentration of CO2 in developing seeds of B. napus (36), which shifts the reaction equilibrium toward carboxylation. Mitochondrial Malic Enzyme—NAD malic enzyme is active in plant mitochondria (37), but flux distributions reported for other plant tissues describe a relatively low contribution by malic enzyme to mitochondrial pyruvate, almost all of which is imported (see Table 2 and references therein). In contrast, 40% of mitochondrial pyruvate is made via malic enzyme in B. napus embryos (Fig. 2 and Table 2). Most Mitochondrial Metabolic Flux Is Devoted to Cytosolic Fatty Acid Elongation—Oil represents the largest reserve of carbon in seeds of B. napus and for the Reston variety 60% of its fatty acids are 20 carbons or more in chain length. While de novo synthesis of C16 and C18 fatty acids takes place in the plastids, elongation to 20 and more carbons takes place in the cytosol (38). The cytosolic Ac-CoA used for elongation of C18 is formed mainly by ATP:citrate lyase (1, 20, 21). As shown in Fig. 2, essentially all the citrate formed in the mitochondria is exported and is used for producing cytosolic Ac-CoA; this represents the largest flux of carbon leaving the mitochondrion. Mitochondrial Metabolism Does Not Provide Precursors for Plastidic Fatty Acid Synthesis—As shown in Fig. 2 (Table 1), about 75% of the pyruvate feeding plastidic fatty acid synthesis is produced by plastidic pyruvate kinase. The remaining 25% of pyruvate in the plastid is imported from the cytosol. Besides these reactions, significant contributions to plastidic fatty acid synthesis from import of cytosolic Ac-CoA or from the generation of pyruvate via plastidic malic enzyme can be excluded (1), and this was confirmed in this study by the labeling signatures in C18(1–2), C22(1–2), and Asp. Balancing of Carbon/Nitrogen Metabolism—The formation of KG as a precursor for glutamate synthesis is regarded as a central function of mitochondrial metabolism. However, as shown in Fig. 2, in B. napus embryos the mitochondrion is a net consumer of KG. A related common feature of cellular anabolism is OAA synthesis via PEP carboxylase with OAA entry into

Mitochondrial Metabolism in B. napus Embryos TABLE 2 Comparison of fluxes related to mitochondrial metabolism for B. napus developing embryos (this study) with published values for different plant and microbial systems CO2 release by mitochondrial reactions Malic enzyme flux (vMEmit)g PEP carboxylase flux (vPEPC)g

B. napus embryosa

Maize root tipsb

Tomato cell culturec

Soy bean embryosd

Escherichia coli e

3.7% 40% 62%

21% 9% 57%

27% 7% 37%

19% 16% 38%

29% 8% 28%

f

a

This study. Dieuaide-Noubhani et al. (18). Exponential phase (Rontein et al. (19)). d Sriram et al. (27). e Glucose fed in bioreactor (Fisher et al. (49)). f Percent of total carbon of hexose catabolized (6 ⫻ vuptGlc ⫺ vST), released as CO2 by ICDH, KDH, PDHmit, and MEmit together (see Table 1). g Percent of flux, relative to pyruvate entry into the mitochondrion (vPDHmit) b c

34046 JOURNAL OF BIOLOGICAL CHEMISTRY

cyclic tricarboxylic acid flux might also be attributed to anaerobiosis in the embryos. In this case fermentative processes could provide ATP instead of oxidative phosphorylation. However, we conclude that anaerobiosis is not the case because 1) we could not detect products of fermentative metabolism in the growth media (unpublished results) nor is a high CO2 evolution rate observed (16, 17); 2) culturing B. napus embryos at increased oxygen levels (60% v/v) did not increase the carbon use efficiency or incorporation of precursors into oil, i.e. oxygen was not limiting the transformation of carbon into storage compounds (17); 3) under the experimental growth conditions photosynthesis is generating oxygen inside the embryo; and 4) although in planta the oxygen tension inside developing seeds of B. napus has been reported to be low, the seeds do not enter anaerobiosis (23, 47). Using a similar steady-state labeling approach as in this study, Sriram et al. (27) studied developing soybean embryos using U-13C-labeled sucrose and unlabeled Gln as sole carbon sources. In comparison with the B. napus embryos of our study, the soybean embryos produce and store much more protein and less oil. Also cytosolic elongation of fatty acids to C20 and C22 chain length is essentially not present in soybean due to the different fatty acid composition of the seed oil. Similar to our results and as a consequence of using Gln as nitrogen source, the soy flux map does not show net anaplerotic flux of OAA into the tricarboxylic acid cycle (27). PEP carboxylase appears to produce only the amount of OAA needed for production of Asp and derived amino acids. In addition, there is no indication of significant glyoxylate shunt activity in the developing soy embryos. In contrast to B. napus, the export of mitochondrial acetyl-CoA via citrate synthase/ATP:citrate lyase, attributed to cytosolic fatty acid elongation, was not observed in soy (27). Also the relative amount of carbon oxidized by activities of the tricarboxylic acid cycle enzymes to CO2 is substantially higher in soy (Table 2), suggesting that in soy embryos the fraction of cellular ATP produced via mitochondrial tricarboxylic acid cycle activity and oxidative phosphorylation is much higher than in the B. napus embryos. This current study documents the broad plasticity of plant mitochondrial and related metabolism and how this metabolism is tailored toward the oil and protein storage functions that dominate oilseed development. The major differences in flux between B. napus embryos and other plant tissues as shown in Table 2 might reflect different transcript levels of tricarboxylic acid cycle related enzymes. However, a comparison of different VOLUME 281 • NUMBER 45 • NOVEMBER 10, 2006

Downloaded from http://www.jbc.org/ by guest on May 16, 2016

the mitochondrion (anaplerotic flux). In this context, PEP carboxylase is generally considered to exert control over the entry of carbon into tricarboxylic acid-derived amino acid synthesis in leaves (39, 40). However, in B. napus, most of the OAA produced is committed to the synthesis of Asp and amino acids derived from it (Fig. 2), rather than being devoted to replenishing tricarboxylic acid cycle intermediates. Directing more carbon from sugars into protein synthesis by increasing anaplerotic flux via PEP carboxylase has also been proposed as a strategy to increase protein content in seeds (41– 43). However, as shown in Fig. 2, anaplerotic flux from PEP via OAA to KG does not take place. This is because B. napus embryos import nitrogen as amino acids (primarily Gln and Ala), which are not only used directly for protein synthesis but also provide nitrogen via transamination/deamination for the synthesis of other amino acids. Consequently embryo metabolism is provided with the KG derived from Gln (Fig. 2), while conventionally the converse would be assumed. We suggest therefore that for seeds that rely on Gln as a substantial nitrogen source it is unlikely that anaplerotic flux into KG can increase protein content. It rather might reduce the use of Gln as a nitrogen source, thus reducing not increasing protein production. Thus Fig. 2 provides an explanation for the observation that seed protein content depends on the identities and amounts of amino acids supplied by the maternal plant and on the metabolic fate of their carbon skeletons (44, 45). Plasticity of Mitochondrial Metabolism—In summary, based on the flux analysis described above, several aspects of mitochondrial metabolism in B. napus embryos are clearly unconventional compared with flux modes previously described. These differences emphasize the highly plastic nature of plant mitochondrial metabolism. The absence of respiratory tricarboxylic acid cycle flux is at first surprising, considering the high biosynthetic ATP demands in an oilseed. However, to assess the need for mitochondrial oxidation and ATP production, all cellular ATP producing and consuming reactions have to be balanced including photosynthetic ATP and reductant production. Ruuska et al. (12) calculated that light absorbed by embryos can provide sufficient ATP and reductant for all oil synthesis. The lack of tricarboxylic acid cycle flux found in B. napus embryos is also consistent with previously described regulatory mechanisms. For example, it has been reported that in leaves the tricarboxylic acid cycle is inhibited by light (46), and embryos may respond similarly to the low light levels that reach seeds growing in planta (9). The observed absence of

Mitochondrial Metabolism in B. napus Embryos tissues in A. thaliana (leaves, roots, seeds) that have different metabolic functions of mitochondrial metabolism indicated comparatively moderate differences in gene expression (48). Therefore, it is likely that plants achieve substantial plasticity of mitochondrial central metabolism via post transcriptional mechanisms in addition to differences in gene expression. REFERENCES

NOVEMBER 10, 2006 • VOLUME 281 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY

34047

Downloaded from http://www.jbc.org/ by guest on May 16, 2016

1. Schwender, J., and Ohlrogge, J. B. (2002) Plant Physiol. 130, 347–361 2. Singal, H. R., Sheoran, I. S., and Singh, R. (1987) Plant Physiol. 83, 1043–1047 3. Murphy, D. J., and Cummis, I. (1989) J. Plant Physiol. 135, 63– 69 4. Kang, F., and Rawsthorne, S. (1994) Plant J. 6, 795– 805 5. Singal, H. R., Talwar, G., Dua, A., and Singh, R. (1995) J. Biosci. 20, 49 –58 6. Eastmond, P., Kolacna, L., and Rawsthorne, S. (1996) J. Exp. Bot. 47, 1763–1769 7. Asokanthan, P. S., Johnson, R. W., Griffith, M., and Krol, M. (1997) Physiol. Plant. 101, 353–360 8. Gupta, R., and Singh, R. (1997) Indian J. Biochem. Biophys. 34, 288 –295 9. King, S. P., Badger, M. R., and Furbank, R. T. (1998) Aust. J. Plant Physiol. 25, 377–386 10. Eastmond, P. J., and Rawsthorne, S. (1998) J. Exp. Bot. 49, 1105–1111 11. Eastmond, P. J., and Rawsthorne, S. (2000) Plant Physiol. 122, 767–774 12. Ruuska, S. A., Schwender, J., and Ohlrogge, J. B. (2004) Plant Physiol. 136, 2700 –2709 13. Kubis, S. E., Pike, M. J., Everett, C. J., Hill, L. M., and Rawsthorne, S. (2004) J. Exp. Bot. 55, 1455–1462 14. Schwender, J., Ohlrogge, J. B., and Shachar-Hill, Y. (2003) J. Biol. Chem. 278, 29442–29453 15. Schwender, J., Ohlrogge, J. B., and Shachar-Hill, Y. (2004) Curr. Opin. Plant Biol. 7, 309 –317 16. Schwender, J., Goffman, F., Ohlrogge, J. B., and Shachar-Hill, Y. (2004) Nature 432, 779 –782 17. Goffman, F. D., Alonso, A. P., Schwender, J., Ohlrogge, J. B., and ShacharHill, Y. (2005) Plant Physiol. 138, 2269 –2279 18. Dieuaide-Noubhani, M., Raffard, G., Canioni, P., Pradet, A., and Raymond, P. (1995) J. Biol. Chem. 270, 13147–13159 19. Rontein, D., Dieuaide-Noubhani, M., Dufourc, E. J., Raymond, P., and Rolin, D. (2002) J. Biol. Chem. 46, 43948 – 43960 20. Fatland, B. L., Ke, J., Anderson, M. D., Mentzen, W. I., Cui, L. W., Allred, C. C., Johnston, J. L., Nikolau, B. J., and Wurtele, E. S. (2002) Plant Physiol. 130, 740 –756 21. Fatland, B. L., Nikolau, B. J., and Wurtele, E. S. (2005) Plant Cell 17, 182–203 22. Kro¨mer, S. (1995) Annu. Rev. Plant Physiol. Plant Mol. Biol. 46, 45–70 23. Vigeolas. H., van Dongen, J. T., Waldeck, W., Hu¨hn, D., and Geigenberger, P. (2003) Plant Physiol. 133, 2048 –2060 24. Mollney, M., Wiechert, W., Kownatzki, D., and de Graaf, A. A. (1999) Biotechnol. Bioeng. 66, 86 –103

25. Wiechert, W., Mollney, M., Petersen, S., and de Graf, A. A. (2001) Metab. Eng. 3, 265–283 26. Galvez, S., and Galdal, P. (1995) Plant Sci. 105, 1–14 27. Sriram, G., Fulton, D. B., Iyer, V. V., Peterson, J. M., Zhou, R., Westgate, M. E., Spalding, M. H., and Shanks, J. V. (2004) Plant Physiol. 136, 3043–3057 28. Salon, C., Raymond, P., and Pradet, A. (1988) J. Biol. Chem. 263, 12278 –12287 29. Edwards, S., Nguyen, B. T., Do, B., and Roberts, J. K. M. (1998) Plant Physiol. 116, 1073–1081 30. Bowsher, C. G., and Tobin, A. K. (2001) J. Exp. Bot. 52, 513–527 31. Ap Rees, T. (1990) in Plant Physiology, Biochemistry and Molecular Biology (Dennis, D. T., and Turpin, D. H., eds) pp. 106 –123, Longman Scientific & Technical, Harlow, UK 32. Dieuaide-Noubhani, M., Canioni, P., and Raymond, P. (1997) Plant Physiol. 115, 1505–1513 33. Stefanopoulos, G. N., Aristidou, A. A., and Nielsen, J. (1998) Metabolic Engineering: Principles and Methodologies, p. 69, Academic Press, London 34. Des Rosiers, C., Fernandez, C. A., David, F., and Brunengraber, H. (1994) J. Biol. Chem. 269, 27179 –27182 35. Des Rosiers, C., Didonato, L., Comte, B., Laplante, A., Marcoux, C., David, F., Fernandez, C. A., and Brunengraber, H. (1995) J. Biol. Chem. 270, 10027–10036 36. Goffman, F. D., Ruckle, M., Ohlrogge, J. B., and Shachar-Hill, Y. (2004) Plant Physiol. Biochem. 42, 703–708 37. Douce, R., and Neuberger, M. (1989) Annu. Rev. Plant Physiol. Plant Mol. Biol. 40, 371– 414 38. Bao, X., Pollard, M., and Ohlrogge, J. (1998) Plant Physiol. 118, 183–190 39. Chollet, R., Vidal, J., and O’Leary, M. H. (1996) Annu. Rev. Plant Physiol. Plant Mol. Biol. 47, 273–298 40. Smith, C. R., Knowles, V. L., and Plaxton, W. C. (2000) Eur. J. Biochem. 267, 4477– 4485 41. Sugimoto, T., Tanaka, K., Monma, M., Kawamura, Y., and Saio, K. (1989) Agric. Biol. Chem. 53, 885– 887 42. Smith, A. J., Rinne, R. W., and Seif, R. D. (1989) Crop Sci. 29, 349 –353 43. Rolletschek, H., Borisjuk, L., Radchuk, R., Miranda, M., Heim, M., Wobus, U., and Weber, H. (2004) Plant Biotechnol. J. 2, 211–219 44. Rolletschek, H., Hosein, F., Miranda, M., Heim, U., Gotz, K. P., Schlereth, A., Borisjuk, L., Saalbach, I., Wobus, U., and Weber, H. (2005) Plant Physiol. 137, 1236 –1249 45. Hernandez-Sebastia, C., Marsolais, F., Saravitz, C., Israel, D., Dewey, R. E., and Huber, S. C. (2005) J. Exp. Bot. 56, 1951–1963 46. Tcherkez, G., Cornic, G., Bligny, R., Gout, E., and Ghashghaie, J. (2005) Plant Physiol. 138, 1596 –1606 47. Geigenberger, P. (2003) Curr. Opin. Plant Biol. 6, 247–256 48. Schmid, M., Davison, T. S., Henz, S. R., Pape, U. J., Demar, M., Vingron, M. Scho¨lkopf, B. Weigel, D., and Lohmann, J. (2005) Nat. Genet. 37, 501–506 49. Fischer, E., Zamboni, N., and Sauer, U. (2004) Anal. Biochem. 325, 308 –316

Schwender et al.: Supplemental Text

Mitochondrial Metabolism in Developing Embryos of Brassica napus Jörg Schwender, Yair Shachar-Hill and John Ohlrogge

Supplemental Text 1. Metabolic Network 1.1 Network definition 1.2 Isotopic imprinting in fatty acids and proteinogenic amino acids 1.3 Subcellular localization of amino acid synthesis 1.4 Subcellular compartmentation of acetyl-CoA

2 4 4 5

2. Isotopomer Model 2.1 Steady state isotopomer model 2.2 Metabolic model combining three labeling experiments

6 6

3. Measurements 3.1 Biomass derived fluxes 3.2 Label measurements

9 9

4. Flux parameter fitting 4.1 General least squares fitting procedure 11 4.2 Computational flux fitting 11 4.3 Complementary labeling approach resolves flux in sub-compartmented cells 13

5. Statistical Analysis 5.1 Error in biomass proportions 5.2 Error in MS measurements

15 15

6. Model Validation 6.1 Isotopic steady state 6.2 Metabolic steady state 6.3 Isotopic equilibrium of transaminases 6.4 Isotopic equilibrium of C4 dicarboxylic acids 6.5 Isotopic equilibrium of cytosolic and plastidic phosphoenol pyruvate 6.6 Pyruvate carrier 6.7 Uptake ratio of Ala and Gln independently confirmed by 15N labeling 6.8 Influence of Atmospheric CO2

17 17 18 18 18 19 19 19

7. Supplemental Tables

21

8. References

32

1

Schwender et al.: Supplemental Text

1. Metabolic Network 1.1 Network definition. A network of central metabolism containing the reactions of glycolysis, oxidative pentosephosphate pathway (OPPP), RubisCO, the tricarboxylic acid (TCA) cycle and fatty acid synthesis was defined as shown in Fig. S1. The key features are the same as reported in Schwender et al. (2003) for B. napus embryos with the addition of the reactions of the TCA cycle as well as the RubisCO and phosphoribulokinase reactions. The core TCA cycle enzymes as well as mitochondrial malic enzyme were considered (Fig. S1). Reactions specific for the glyoxylate cycle were not considered because significant activity of the key enzymes have not been detected during the main phase of storage accumulation (Chia et al., 2005). In fact, if mitochondrial isocitrate lyase together with malate synthase were added to the model, flux parameter fitting consistently assigned very small fluxes to these reactions and the quality of the fit was not increased by their inclusion (data not shown). In relation to the TCA cycle, cytosolic ATP:citrate lyase was considered to produce cytosolic acetyl-CoA for the cytosolic elongation of fatty acids as described before (Schwender and Ohlrogge, 2002). Also the RubisCO reaction was added as this enzyme is known to be abundant and active in vivo in developing B. napus embryos (Ruuska et al., 2004) and has been shown to carry flux in vivo (Schwender et al., 2004). The network is shown compartmentalized only in reactions downstream of phosphoenol pyruvate because labeling experiments indicated a high exchange of several glycolytic intermediates between cytosol and plastid (Schwender et al., 2003). Therefore, reactions upstream of PEP are modeled without considering separation into cytosolic and plastidic compartments. For example, vG6PDH summarizes the flux through both plastidic and cytosolic G6PDH.

2

Schwender et al.: Supplemental Text Starch Glc6PDH

Glc

CO2

6PGlcnDH

6-PGlcn

HK

S-1,7P2

Glc-6P

Sucrose

GPI Susy

Fru-6P PFK

S-7P

Riso

Ru-5P

TK

Xu-5P

E-4P

TK

R-5P

ShBPase

SAdlo

Xepi

FBPase

PRK

Fru-1,6P2 His, Trp Phe, Tyr, Trp

Aldo TPI

DHAP

RuBisCO

GAPDH, PGK

Glycerol in TAG

Ru-1,5-P2

GAP

PGA

CO2

PGM, Eno

CO2 PEP

Ala

Ala

PK

PK

PEPC AAT

Phe, Tyr, Trp

PEP

Pyr

Val, Leu, Ile, Lys

Pyr

PDH

Ala

CO2

ACL

Pyr PDH

Asp,Thr, Met, Ile, Lys OAA

ME

CS

AcCoA

Cit

OAA

Aco

MDH

Mal

Mitoch. KDH ScK

ScDH

ICit

KG

Leu

Ac-CoA ACC, FAS

ACE

ICDH

Fm

Fum

CO2

Cit

CO2 AcCoA

FA (C16, C18)

CO2 Plastid

Succ CO2

Gln

Gln Gln

GS

GAT

Glu

Cytosol VLCFA (C20, C22)

Glu, Pro, Arg

TAG

Fig. S1 Metabolic network, showing the main reactions of central metabolism known to be present in developing, oil accumulating B. napus embryos. The carbon sources provided to embryo cultures are shown as well as all the major biosynthetic products stored in the seeds (red). Dashed boxes designate central metabolites that were assumed to equilibrate isotopically by fast interconversion based on experimental data (see section 6) Enzymes: ACC: acetyl-CoA carboxylase; ACE: acyl-CoA elongase; ACL: ATP:citrate lyase; Aco: aconitase; Aldo: fructose 1,6-bisphosphate aldolase; AAT: alanine transaminase; CS: citrate synthase; Eno: enolase; FBPase: fructose 1,6 bisphosphatase; FAS: fatty acid synthase; Fm: fumarase; GAPDH: glyceraldehyde 3-phosphate dehydrogenase; GAT: glutamyl aminotransferase; Glc6PDH: glucose 6-phosphate dehydrogenase; GPI: glucose 6-phosphate isomerase; GS: glutamate synthase; HK: hexokinase; ICDH: isocitrate dehydrogenase; KDH: ketoglutarate dehydrogenase; MDH: malate dehydrogenase; ME malic enzyme; PDH: pyruvate dehydrogenase; PEPC: phosphoenol pyruvate carboxylase; PFK: phospho fructokinase; PGM: phosphoglycerate mutase; PGK: phosphoglycerate kinase; PK: pyruvate kinase; PRK: phosphoribulokinase; 6PGlcnDH: 6-phosphogluconate dehydrogenase; Riso: ribose 5-phosphate isomerase; RuBisCO: ribulose 1,5-bisphosphate carboxylase/oxygenase; SAldo: sedoheptulose 1,7-bisphosphate aldolase; ScDH: succinate dehydrogenase; ScK: succinate kinase; ShBPase: sedoheptulose bisphosphatase; SuSy: sucrose synthase; TK: transketolase; Xepi: xylulose 5-phosphate epimerase; TPI: triose phosphate isomerase. Metabolites: 6-PGlcn: 6-phosphogluconate; Cit: citrate; DHAP: dihydroxy acetone phosphate; E-4P: erythrose 4-phosphate; Fru-1,6P2: fructose 1,6-bisphosphate; Fru-6P: fructose 6-phosphate; Fum: fumarate; GAP: glyceraldehyde 3phosphate; Glc-6P: glucose 6-phosphate; ICit: isocitrate; KG: ketoglutarate; Mal: malate; OAA: oxaloacetate; PEP: phosphoenol pyruvate; PGA: 3-phosphoglycerate; Pyr: pyruvate; R-5P: ribose 5-phosphate; Ru-1,5P2: ribulose 1,5-bisphosphate; Ru-5P: ribulose 5-phosphate; Sh-1,7P2: sedoheptulose 1,7-bisphosphate; Sh-7P: sedoheptulose 7-phosphate; Succ: succinate; TAG: triacyl glycerol; VLCFA: very long chain fatty acids; Xu-5P: xylulose 5-phosphate;

3

Schwender et al.: Supplemental Text

1.2 Isotopic imprinting in proteinogenic amino acids and fatty acids. In a retro-biosynthetic approach the labeling in central metabolites such as phosphoenol pyruvate, pyruvate or oxaloacetate can be derived from the labeling of their products Phe, Val and Asp, respectively (Szypersky 1998). The particular amino-acid–precursor relations are summarized in Table S4 and Fig. S2. Instead of extracting and analyzing labeling in low abundance intermediates, labeling in storage proteins and oil was measured and analyzed by GC/MS methods as described before (Schwender and Ohlrogge 2002, Schwender et al., 2003). The fractional 13C-enrichment in protein amino acids and fatty acids was interpreted according to the biosynthetic relationship between precursors and their products accumulating in biomass (Fig. S2). 1.3 Subcellular localization of amino acid synthesis The following assumptions were made about the subcellular localization of amino acid syntheses: The biosyntheses of His, Val, Leu, and Ile are exclusively plastidic (Ohta et al., 2000; Singh et al., 1999). Their labeling represents that in plastidic pentose phosphate (His), pyruvate (Val) and aspartate (Leu and Ile). Also the key enzymes in aromatic amino acid (Phe, Tyr, Trp) biosynthesis are known to be plastid-localized. In the absence of photorespiration in B. napus embryos (Schwender et al. 2002, Goffman et al., 2004), serine is formed from PGA by the plastidic phosphorylated serine biosynthetic pathway (Ho et al., 1999). However, a signature of the interconversion of Ser and Gly by serine hydroxymethyl transferase can be observed in the labeling pattern of Ser (data not shown) making Ser unsuitable as a reporter for labeling in PGA. Protein amino acids were assumed to be derived from their cytosolic pools independent from the subcellular compartment of their biosynthesis. This is because the protein analyzed consists almost entirely of storage proteins which are synthesized in the cytosol, presumably from amino acid tRNAs formed in the cytosol. Aspartate and Glutamate can be derived from oxaloacetate and α-ketoglutarate, respectively, in different compartments by transamination (Schultz and Coruzzi, 1995; Wilkie and Warren, 1998). Assuming abundant cytosolic aminotransferase activities, one can assume that protein-bound Asp and Glu represent cytosolic OAA and KG, respectively.

4

Schwender et al.: Supplemental Text 1

1

Ala(1-3)

Ala(2-3)

Pyrcyt

Alaprotein

1

1

Asp(1-4)

Asp(1-2) Asp(2-4)

OAAcyt

Aspprotein

1 2 3

Val(1-5)

Pyrpl

5 4 3 2 1

His(1-6) His(2-6)

R-5Ppl “C1” pl

1

1 2 3 4 5

1

1 2 3 4 5

Pro(1-5) Pro(2-5)

KGmit

Glu/Procyt Proprotein

Glucose(1-6)

1

Glucose(1-2) Glucose(4-6)

N

HPpl

Glucosestarch

1 1

TPcyt

Val(2-5)

3

N

Hisprotein

glycerol(1-3)

3

Valprotein

1 1 2 3 4 5

1 2 2

glycerol(2-3)

GlycerolTAG (symmetric)

C18(1-2)

1

AcCoApl

1

C22(1-2)

1

AcCoAcyt

C18TAG

C22TAG

Fig. S2 Relationships between particular carbon atoms of central metabolites and those of monomers in biomass compounds. The numbers in the carbon atoms (circles) denote the carbon positions of the precursors while the indices (subscripts) defining the MS fragments are the carbon positions of the monomers. The carbon transitions of amino acid biosynthesis can be derived from general biochemistry textbooks.

Since protein is acid-hydrolyzed prior to amino acid analysis Gln is deaminated to Glu. Therefore Glu in the hydrolysate is derived from both Gln and Glu of the protein. Therefore, the label observed in protein-derived Glu was not used for modeling. Instead Pro, which is directly biosynthetically derived from Glu, was used and assumed to represent the label in cytosolic Glu. In several experiments, when labeled free amino acids were extracted and analyzed, the labeling pattern in Glu and Pro were always found to be identical, showing labeling signatures related to mitochondrial ketoglutarate (data not shown). 1.4 Subcellular compartmentation of Acetyl-CoA. In plants, fatty acid synthesis is predominantly localized in plastids (Ohlrogge et al., 1979, Dennis, 1989). Plastidic fatty acid synthesis produces C16 and C18 fatty acids, whereas the elongation of C18:1 by a cytosolic fatty acid elongation system produces C20 and C22 fatty acids (Domergue et al., 1999; Whitfield et al., 1993). Thus, labeling in the carboxyl-terminal acetate units of C18 and C22 fatty acids represent plastidic and cytosolic acetyl-CoA pools, respectively (Schwender and Ohlrogge, 2002). Terminal acetate units of C18 and C22 methyl esters are monitored in the Fragment m/z 74 (McLafferty fragment) as described in Schwender et al. (2003).

5

Schwender et al.: Supplemental Text

2. Isotopomer Model A steady state metabolic model was used in this study with the topology of the metabolic network defined by the stoichiometries and the carbon transitions of the biochemical reactions. The biomass composition of the embryos defines the fluxes of metabolites into biomass. Given the 13C-labeling of substrates and values for the fluxes, the model simulates all MS measurements. The model described here is able to simultaneously simulate MS measurements for three different labeled substrates ([1,2-13C2]glucose/[U13 C6]glucose, [U-13C5]Gln and [U-13C3]Ala, respectively). 2.1 Steady state isotopomer model. Isotopomer balancing and flux parameter fitting were performed using the software package 13CFLUX (obtained from Prof. Dr. W. Wiechert, Department of Simulation, University of Siegen, Germany) on a X86 compatible computer as described by Wiechert and co-workers (Wiechert et al. 2001, and refs cited therein). The software was run under RedHat Linux. 13CFLUX allows the implementation of metabolic network models by allowing the user to define the label input, the network stoichiometry, the carbon transitions and definitions of measured labeling patterns (NMR and MS data). Biochemical conversions are defined as net and exchange fluxes (Vnet, VXch) which are related to the conventional forward and reverse fluxes according to Wiechert and DeGraaf, 1997: Vforward = Vnet + VXCH Vbackward = Vnet – VXCH In cases where a reaction is assumed to be irreversible, the respective exchange flux is set to zero and the value of the flux is constrained to being positive. With 13CFLUX it is possible to fit simulated labeling patterns to experimental data (Non-linear least-squares fitting approach). Several evolutionary algorithms of the software-package were used to find flux values that most likely represent a global optimum fit (section 4). Statistical tools in 13CFLUX allow the errors in the fluxes to be derived based on linearized statistical analysis (see section 5). 2.2 Metabolic model combining three labeling experiments. The central metabolic network for developing B. napus embryos (Fig. S1) containing the reactions of glycolysis, oxidative and non-oxidative pentosephosphate pathway (PPP), RubisCO, the tricarboxylic acid (TCA) cycle and fatty acid synthesis was implemented by defining stoichiometry and carbon transitions, which are taken from standard biochemistry textbooks. An overview over all metabolic pools and reactions implemented in the model is given in Fig. S3. Essential parts of the 13CFLUX model files for embryos are shown in Table S10.

6

Schwender et al.: Supplemental Text Glc0

GlcU

vupt0

vuptU Glc

VuptGlc

vHP_St

vME

HP

St

vG6PDH vRub vKDH

vCS

R-5P Xu-5P

Phe, Tyr, Trp)

Riso

Ru-5P Xepi

vTK2

vPDH vCO2_out

EP vSaldo SP

vAldo

CO2_out

½ vVal vICDH

vHis_P

vG6PDH

(Starch, cell wall)

vPEPC

CO2

Aro (R5P into

CO2

vTK1

vEP_P

Aro (E4P into CO2

glycerol

vGlyc_TAG

(in TAG)

TP

Phe, Tyr, Trp)

vRub

vGAPDH vSer_P

SerP

PGA

(Ser, Gly)

vPGM CO2 c EP vP

PEP_P (PEP into Phe, Tyr, Trp) ValP (Pyr into Val, Lei, Ile, Lys)

vPKpl

PEP vPKc

vuptAla Ala_in

P_P vPE

vAAT

AlaCyt

vPyrpl_P vPyr1

PyrCyt

PyrPl

Val CO2

vAla_P

vPyr2

CO2 vPDHpl vAcpl_P Acpl AcPl

AlaP vC S

CO2 vME

OAA vAsp_P

vFASpl

CO2

Cit

[Ac-CoA]

Pyrmit

vACL vAco

vFMa vFMb AspP (Asn, Thr, Met, Ile, Lys)

Succ

AcCyt

vFAScyt

Icit

AcFA

vFAS

vICDH

vKDHa vKDHb

CO2

KG

AcFA_out (total fatty acid synthesis)

CO2 vGAT Glu Gln vuptGln Gln_in

vGlnGlu vGluP GluP

vGlnP

Free Net fluxes Free exchange fluxes

xxx

Biomass derived Net fluxes

xxx

MS Measurements Input label

(Glu, Pro, Arg)

GlnP

Fig. S3: Model network as derived from Fig. S1 and implemented in 13CFLUX (See Table S10). The figure shows the reactions with flux and metabolite names used in the model file. Input metabolites are shown in green. Free variable fluxes are in blue. Fluxes into biomass synthesis (biomass constraints) are in red. Double headed arrows designate reactions that are modeled reversibly. Ac: Acetyl-CoA; AcFA: acetate units in fatty acids; Cit: citrate; ICit: isocitrate; EP: erythorse 4-phosphate; HP: hexose phosphates; PDH: pyruvate dehydrogenase; PEP:phosphor enolpyruvate; PGA: phosphoglycerate; Pyr: pyruvate; PK: pyruvate kinase; R-5P: ribose 5-phosphate; Ru-5P: ribulose 5-phosphate; SP: sedoheptulose 7-phosphate; TP: triose phosphates; Xu-5P: xylulose 5-phosphate.

In order to simultaneously predict labeling measurements for experiments with different labeled substrates in one model the 13CFLUX model file contains a repetition of all reaction equations of the basic flux model shown in Fig. S3. The parallel sub modelnetworks were distinguished by their flux names (see Table S10). In each sub-network a flux name () was designated according to “v”, “v2” or “v3” and metabolite names accordingly as “m1”, “m2” and “m3”. Also, different definitions for label input metabolites in the three sub-networks were defined according to Fig. S4A.

7

Schwender et al.: Supplemental Text The sub models were connected by adding linear flux dependencies (section “EQUALITIES” in the 13CFLUX model file) in a way that defines all the fluxes of the second and the third sub-models to equal the corresponding fluxes in the first one (Fluxes defined as dependent). In this way, by definition of the free fluxes in the first sub-model, the complete model simulates the isotopomer distributions of all three labeling experiments dependent on the flux distribution in the model shown in Fig. S3. This allows the optimization of flux parameters to simultaneously fit the labeling data of all three experiments using different labeled precursors.

A

Medium composition Exp. 1

Exp. 2

Exp. 3

Sucrose unlabeled

unlabeled

unlabeled

Glucose [1,2-13C2]Glc / [U-13C6]Glucose (1:1)

unlabeled

unlabeled

Ala

unlabeled

[U-13C3]Ala

unlabeled

Gln

unlabeled

unlabeled

[U-13C5]Gln: Gln 1:1

Measured fragments, mass isotopomers

0.3

0.2

0.1

0.0

0.3

Asp(1-2)

0.2

Asp(1-4)

0.1

Val(2-5)

0.0

Val(1-5)

0.3

Acpl(1-2)

0.2

Alac(2-3)

0.1

0.0

B

Alac(1-3)

m0 m1 m2 m3 m0 m1 m2 m0 m1 m2 m0 m1 m2 m3 m4 m5 m0 m1 m2 m3 m4 m0 m1 m2 m3 m4 m0 m1 m2

m0 m1 m2 m3 m0 Acc(1-2) m1 m2 m0 m1 m2 Pro(1-5) m3 m4 m5 m0 m1 Pro(2-5) m2 m3 m4

Asp(2-4)

Fig. S4: Combination of three labeling experiments in flux parameter fitting and statistical analysis (see also table S6). A) Medium composition in the three labeling experiments. B) Fractional 13Cenrichments (mass isotopomer abundance) in 11 selected fragments resulting from the three experiments. As shown, each experiment results in a different labeling pattern for most fragments.

8

Schwender et al.: Supplemental Text

3. Measurements The flux estimation was constrained by the 13C-labeling pattern in metabolites and by the biomass composition which together allow fluxes of central metabolism and seed storage product synthesis to be derived. In addition 15N labeling data were used to derive the extent of reversibility of transaminase reactions as described below under sections 4.1 and 6.3. 3.1 Biomass derived fluxes. Fluxes of metabolites (e.g. pyruvate or oxaloacetate) out of central metabolism into oil, storage protein and carbohydrate polymers (Fig. S3, red arrows) are derived from the biomass composition. For embryos grown in culture, oil, protein, starch, and cell wall polymers were considered as biomass components (Table S1). From the fatty acid composition of seed oil (Table S3) and the amino acid composition in storage protein (Table S2) the quantities of monomers (proteinogenic amino acids as well as glycerol and cytosolic and plastidic acetyl units, respectively) can be derived. The amount of hexose-phosphate used for glucose polymers (starch and cell walls) comes directly from the molar weight of glucose in a glucose polymer ((180 – 18) g/mol). With the stoichiometries of the biosyntheses of the monomers, the biosynthetic demands on the central metabolites (like pyruvate, OAA) can be derived. Finally, by considering the growth rate of the embryos an absolute flux rate “[µmol h-1 gDW-1]” of intermediates into biomass synthesis can be given as shown in table 1 of the main text. 3.2 Label Measurements. As described in the main text, the measured mass isotopomer fractions are averages from three GC/MS runs and were corrected for the contributions of natural abundance of isotopes of C, H, N, O, S and Si. For a given Cn-fragment in the mass spectrum of a labeled compound the corrected mass isotopomer fractions are [m0, m1, … mn] with the sum of all mass isotopomers equal to 1 or 100 %. In the following the resulting corrected mass isotopomer fractions are called “Label Measurements”. The label measurements were defined in the 13CFLUX model files as “cumomer fractions” (Wiechert et al., 1999). For each of the three labeling experiments more than 150 mass isotopomer fractions were determined by GC/MS analysis. Out of these data in each experiment 47 label measurements were added into the flux model (see Fig. S4, Table S6) (TBDMS derivatives of Asp, Pro, Ala, Val; C18 and C22 fatty acid methyl esters). Not used in the flux model were measurements of amino acids which have the same precursor (e.g. Thr is derived from Asp and always showed the same labeling pattern as Asp). In the case of the [1,2-13C2]glucose / [U-13C6]glucose labeling, MS measurements of glucose-methoxime pentaacetate, TBDMS-histidine and the glycerol-TFA were also considered - being derived from starch, protein and triacylglycerols, respectively (Table S7). The labeling in Ribose-5P was derived from the labeling in histidine which is composed of a ribosyl part and N10-Formyl-THF derived carbon (C-6 of His; Fig. S3). Assuming that C-2 of Gly represents labeling in N10-formyl-THF, the labeling in Riblose-5P (RP(1-5)) was derived from the labeling in His (His(1-6)) and Gly (Gly(2)) by linear regression according to:

9

Schwender et al.: Supplemental Text ⎡m0 ⎢m ⎢ 1 ⎢0 ⎢ ⎢0 ⎢0 ⎢ ⎢0 ⎢0 ⎣

0

0

0

0

m0

0

0

0

m1 0

m0

0

0

m0

0

0

m1 0

m0

0

0

m1 0

0

0

0

m1 0

0⎤ ⎡h0 ⎤ ⎡r0 ⎤ ⎢ ⎥ ⎥ 0 ⎥ ⎢ ⎥ ⎢ h1 ⎥ r1 0 ⎥ ⎢ ⎥ ⎢h2 ⎥ ⎥ ⎢r2 ⎥ ⎢ ⎥ 0 ⎥ × ⎢ ⎥ = ⎢ h3 ⎥ r3 0 ⎥ ⎢ ⎥ ⎢h4 ⎥ ⎢ ⎥ r4 ⎥ ⎢ ⎥ m0 ⎥ ⎢ ⎥ ⎢ h5 ⎥ ⎢ r5 ⎥ m1 ⎥⎦ ⎣ ⎦ ⎢⎣h6 ⎥⎦

Mass isotopomer mapping matrix Derived from Gly(2).

RP(1-5) His(1-6)

(In the mapping matrix m0 and m1 describe the fractional 13C-label in C2 of glycine, as derived by MS measurements from TBDS-Gly (Gly(2)). Vector RP(1-5) [r0 … r5]: fractional 13C-label in RP(1-5); Vector His(113 6) [h0 … h6]: fractional C-label in His(1-6) as derived from MS measurements of TBDS-His.)

10

Schwender et al.: Supplemental Text

4. Flux parameter fitting 4.1 Least squares fitting procedure: In order to find flux values in the model network that best explain the biomass derived fluxes and fractional labeling data, the simulated MS measurements calculated by the model are compared to the experimental data. The problem of finding the best fit is solved by a non-linear least-squares fitting approach. The procedure of flux parameter fitting is summarized in Fig. S5. The sum of squared differences between model predictions and experimental measurements is minimized by variation in the free flux parameters. Due to the non-linearities in the isotopomer model there is no guarantee that an identified minimum of least squares represents the global minimum. Therefore, as is general practice in literature on 13C-flux analysis the flux values obtained should be referred to as the “best estimates”.

Label input

Biomass derived Fluxes

1) [1,2-13C2]/[U-13C6]Glucose 2) [U-13C3]Ala 3) [U-13C5]Gln

Growth rate

vFAScyt

vPyrpl_P

vFAS

vAsp_P

vGlyc_TAG

Starch

Isotopomer model

vHP_St

Three parallel networks (1), 2), 3)) Predicted MS Measurements pi

vGln_P vPEP_P vAcpl_P vHis_P

adjust free fluxes

vGlu_P

xi

Compare prediction with measurements ei = xi - pi n

χ2 = ∑ i =1

Optimizer (CooolEvoAlpha, Donlp2)

vAla_P

St ar tv al ue s

Fatty acid composition

Calculate all dependent fluxes Test all flux constraints

Set of feasible free fluxes

Biomass proportions: Protein Oil Starch Amino acid composition

Random values for 18 free fluxes

ei2

δ i2

vEP_P

Experimental MS data n = 3 x 47 MS measurements (Table S6) 1) [1,2-13C2]/[U-13C6]glucose 2) [U-13C3]Ala 3) [U-13C5]Gln + 23 Additional measurements for exp. 1) (see Table S7)

Minimal χ2

vSer_P

Best estimate for free fluxes

Fig. S5: The procedure used for flux parameter fitting. Details see text. For flux names compare Fig. S3. ei : residuals; xi : measurements; pi: simulated values; δi : standard deviation of measurement i.

4.2 Computational flux fitting. For flux optimization, the program CooolEvoAlpha, an evolutionary algorithm from the 13CFLUX software package as well as the 13CFLUX implementation of Donlp2 (sequential quadratic programming algorithm by Peter Spellucci, Technische Universität Darmstadt, Germany) were used. Both algorithms gave the same general results while Donlp2 needs considerably less computing time. During optimization, the sum of errors (squared differences between model predictions and experimental measurements, weighted by SD2) is minimized (Fig. S5). The algorithms continue variation of the flux parameters until the sum of errors cannot be further diminished. Wide-ranging tests have shown that whether the flux parameter fitting is performed with the original SD derived from MS measurements or by assuming one

11

Schwender et al.: Supplemental Text value for all standard deviations (see section 5), the resulting best estimate of flux is consistently reproduced and is virtually the same. Flux optimization was repeated with random generated flux parameter sets as starting values. Only those flux sets were used which are feasible within the stoichiometric constraints in the flux parameter space of the network. Initially, analysis of 30 optimized flux sets showed that the algorithm consistently converged to very similar low residuals with one solution for most free net fluxes, while vPyr1 (see Fig. S3) was fixed with some variation. Within this variability vPyr1 is negatively correlated with the exchange flux vAATXCH, which describes the interconversion rate of Pyrcyt and Alacyt (Alanine amino transaminase). This means the model explains the MS data equally well with different combinations vAATXCH and vPyr1. This suggests that the part of the network connecting cytosolic alanine and plastidic pyruvate, in particular the exchange flux vAATXCH, cannot be resolved by 13C-data alone. This missing information is contained in data resulting from labeling with 15N-Ala or 15 N-Gln (table S7a) because reversible transaminase activity exchanges amino groups between amino acids. Assuming that the α-nitrogen of Ala is derived from alanine which is derived from the medium and labeled at 99 % 15N, as well as derived by transamination from 14N-Glu, the steady state equations for the α-nitrogen of Alac can be derived (Fig. 1 in main text): 1) Mass balance of 15N-label:

(vAlaP + vAATNET + vAATXCH )× [%15 N ( Alac )] = vuptAla × [%15N ( Alamedium )]+ vAATXCH × [%15N (Glu protein )] 2) The 15N enrichment in Alaprotein equals the enrichment in Alac. As well the label in Proprotein represents Glu. 3) steady state dAlac = vuptAla + vAATXCH = vAlaPr ot + vAATNET + vAATXCH = 0 dt From 1), 2) and 3) we obtain:

[ [

] [ ] [

] ]

%15N ( Alamedium ) − %15 N ( Ala protein ) vAATXCH = 15 vupt Ala % N ( Ala protein ) − %15 N (Pr o protein )

Using these equations and the 15N-measurements (table S7a), the value for vAATXCH was constrained to 25.8 ± 11.4 times the input flux of alanine (vuptAla). In this range of values, AlaC and PyrC will always be near isotopic equilibrium. As a result of this consideration the exchange flux vAATXCH was fixed to be 25.8 times the influx of Ala (vuptAla). After the adjustment for vAATXCH based on 15N-labeling results the optimizer algorithm was started 100 times with random start values for the free fluxes (Fig. S3) and the evolutionary algorithm converged consistently to one solution with lowest “residuum” (residual sum of deviations, see figure S5) for the free net fluxes, suggesting a global optimum. The best flux estimates are shown in table S8 and table 1 in the main text. 12

Schwender et al.: Supplemental Text If the model predictions and the MS measurements are compared for the best estimate set of flux values, 13 signals with the highest deviations from their respective measurements and contributing 50 % of the error to the Residuum were selected (Table S6, S7). The biggest deviations were observed for m0 signals, indicating that the effect of dilution of label by unlabeled precursors was not always exactly explained by the model. Possibly a correction for the presence of unlabeled biomass could further improve the fit. If the above 13 measurements were removed from the MS data set and flux parameter fitting started again, the model still consistently converged to the same solution. This means that these measurements are truly redundant and can be removed from the data set without loss of information - and demonstrates the robustness of the model achieved by over-determination. 4.3 Using multiple different labeling experiments improves flux determination in compartmented cells In plant cells, subcellular compartmentation increases the number of metabolic pools and fluxes connecting these pools, which presents a challenge to reliably estimating the fluxes of central metabolism (Schwender et al., 2004b). The most common approach in 13Cconstrained metabolic flux analysis is to use labeled glucose. The information content in the labeled metabolites with respect to flux can be improved by a priori design of labeling experiments, resulting in optimized mixtures of different labeled glucose (Mollney et al., 1999). For example a mixture of [1,2-13C2]glucose and [U-13C6]Glucose can be more informative than using [U-13C6]glucose alone. Here we improve the information content of the labeling experiments by combining labeling data from three different labeling experiments ([1,2-13C2]glucose / [U-13C6]glucose, [U-13C3]Ala, and [U13 C5]Gln). If this study had been performed only using a mixture of [1,2-13C2] and [U13 C6]glucose, the resulting standard deviations in several net fluxes would be more than 3 times higher, rendering vPyr1, vPyr2 and vME not well determined. Also, the model with only [1,2-13C2]/[U-13C6]Glucose measurements did not converge consistently to one solution. There were equally good fits with high variation in the flux vPyr1. The effect of combining labeling data is also illustrated in Fig. S6. There is a low sensitivity for vPyr1 in the MS data after labeling with [1,2-13C2]/[U-13C6]Glucose while the labeling with [U13 C3]Ala contributes most to determine this flux (Fig. S6). Other key fluxes are influenced by measurements from all three labeling experiments i.e. they represent overlapping but not identical information (Fig. S6). In a priori design of labeling experiments, the information content of labeling experiments with different labeled substrates is compared by using the volume of the confidence ellipsoid, a statistical measure derived from the covariance matrix of the free fluxes (Mollney et al., 1999). By the same approach the information content for the three labeling experiments ([1,213 C2]/[U-13C6]Glucose, [U-13C3]Ala and [U-13C5]Gln) was calculated separately and with combined data sets. Accordingly, the average radius of the confidence ellipsoid shrinks at least 3 fold if all three data sets are used in fitting.

13

Schwender et al.: Supplemental Text

Flux

Acpl(1-2) Alac(1-3) Alac(2-3) Acc(1-2) Pro(1-5)

Pro(2-5)

Asp(1-4)

Asp(2-4) Asp(1-2)

m0 m1 m2 m0 m1 m2 m3 m0 m1 m2 m0 m1 m2 m0 m1 m2 m3 m4 m5 m0 m1 m2 m3 m4 m0 m1 m2 m3 m4 m0 m1 m2 m3 m0 m1 m2

vME

vICDH

vICDHXCH

a b c a b c a b c a b c a b c

0.06

a [1,2-1313C2]/

[U- C6]Glucose

b [U-13C3]Ala 0.05 c [U-13C5]Gln 0.04

0.03

0.02

0.01

0

-0.01

dmeasurement / dFlux

MS measurements (Fragment, mass isotopomers)

vPEPC

vPyr1

-0.02

-0.03

-0.04

Fig. S6 Increased information content and flux estimate reliability is obtained by combining experiments with different labeled substrates. The sensitivities of several flux values to the MS measurements are shown. Values derived from the output of the isotopomer model (output sensitivity matrix) are shown in color-coded form. Each row refers to an individual mass isotopomer. Each value (dmeasurements/dFlux) indicates the change in a predicted mass isotopomer with a very small (differential) change in a free flux (measurements in the scale of fractional enrichment; Flux in nmol h-1 mg FW-1). Either dark red or dark blue indicates strong positive or negative correlation between MS measurement and flux as indicated by the scale at the right side. White areas in the figure indicate that these MS measurements do not change with a small change in a particular flux, i.e. these measurements probably do not contain information that helps to define the value of fluxes. The more strong correlations there are, the higher will be the reliability of the flux estimate because redundant measurements exist. One can see that information for the free flux vPyr1 is mainly contained in the experiment with [U-13C3]Ala whereas fluxes for PEPC and ME are best contributed by labeling with [U-13C5]Gln.

14

Schwender et al.: Supplemental Text

5. Statistical Analysis Normally distributed measurement errors were assumed. With the given best flux estimates and based on the errors in the measurements, the errors in fluxes were calculated via “linearized statistics” (Wiechert et al., 1997) using the 13CFLUX tool EstimateStat. As outlined below, in order to get realistic values for the errors in the fluxes, we considered the uncertainties in the biomass composition (5.1) as well as the errors in the MS measurements that are derived from experimental repetition of the embryo cultures (5.2). 5.1 Error in biomass proportions (Tables S1 – S3). For fluxes derived from biomass proportions statistical errors were generated by a Monte Carlo approach. Uncertainties in biomass composition were assumed to have a SD of 10% in protein content, oil content and amino acid composition as well as a 10 % uncertainty in the relative abundance of cruciferin, napin and oleosin in storage protein. The statistical variation in the fatty acid composition is taken from Schwender and Ohlrogge (2002). Using these standard deviations random noise was added to the biomass parameters. The random number generator of Microsoft EXCEL was used and N(µ, δ) distributed numbers were derived according to Box and Müller (1958). Biomass derived fluxes were calculated as described in section 3.1 using the noise added values for biomass composition. The repetition of this procedure 20 times leads to standard deviations in the biomass derived fluxes (Tables S2, S3). 5.2 Error in MS Measurements. In total 164 measurement values derived from GC/MS were considered for flux parameter fitting as well as for the estimation of statistical noise (Fig. S4, Table S6 + S7). Each individual measurement is an average of three GC/MS runs of the same sample. For 90 % of all MS peaks the instrumental SD was < 0.3 % in the fractional enrichment scale. Considering the three 13C-labeling experiments ([1,213 C2]Glc /[U-13C6]Glc, [U-13C5]Gln, [U-13C3]Ala) the real errors associated with biological repetition should be much larger. In labeling experiments with B. napus embryos growing on [U-13C6]glucose and averaging three biological repetitions we observed biological standard errors in MS measurements between about 0 and 2 % (data not shown). Since there are three independent experiments combined in the dataset, the error of biological repetition should be contained in the data, although not as repetitions of each individual mass spectrometric measurement. In the absence of biological repetition for each individual MS measurement we estimated one value as a standard deviation being assigned to all MS measurements in the following way: Under the assumption that the difference between the model-predicted and experimental MS measurements is caused only by standard distributed random noise, the Χ2 distribution predicts the expected sum of errors (errors weighted by the standard deviations). Accordingly, on a 90 % confidence level and with 110 degrees of freedom in the system (see Box 1), the weighted sum of errors should be equal or less than 131 ( Χ2110, 90 % = 131, see Box 2). This is fulfilled if the SD in the MS measurements are 1.0 % or bigger (scale of fractional enrichment; see equ. 2 in Box 2). This means that according to the presence of random noise in the system a SD of at least 1.0 % is to be expected in the MS

15

Schwender et al.: Supplemental Text measurements. The estimate of > 1.0 % falls within the range of our reference labeling experiments that showed that the SD in MS measurements are between 0 and 2 %.

Box 1: Degrees of Freedom in the metabolic network: As shown in Tables S6 and S7, 37 molecule fragments with 164 MS measurements are considered in the flux model. For each molecule fragment the sum of mass isotopomers amounts to 1 (100 %). Therefore, for each fragment one value is always redundant and the 164 measurements are reduced by 37 redundancies. In addition, the lower part of the network is determined by 17 free fluxes. In order to solve the system algebraically an equal number of measurements must be given. Therefore, the number of non-redundant MS measurements minus the 17 free fluxes in the system equals the number of measurements by which the system is over-determined, which amounts to 110 (degrees of freedom). In this calculation we do not consider the biomass derived fluxes since these are not changed in flux parameter fitting and we only want to assess the error in the MS measurements.

Using the 1.0 % estimate for SD in MS measurements, the SDs of the fluxes were derived (table S8; table 1, main text) and all net fluxes appear to be statistically determined (i.e. the error is lower than or in the order of magnitude of the flux value). Box 2: Estimation of standard deviation for MS measurements. With 110 degrees of freedom in the system the random error in the measurements can be expected to be equal or lower than 131, based on a 90 % significance level ( Χ2110, 90 % = 131). If the individual SD associated with n measurements could be calculated the error in the system would be: n

(1)

χ2 = ∑ i =1

ei2

δ i2

,

with ei = xi - pi (n independent measurements, xi being a particular MS measurement and pi the model prediction for that measurement; δi is the SD in each xi). In order to estimate the SD from the residuals (ei), one global δ is assumed in approximation for all measurements. Accordingly equation (1) changes to: (2)

n

χ =∑ 2

i =1

ei2

δ

2

=

1

δ

2

n

∑e i =1

2 i

.

Based on equation (2) and knowing that the real error Χ2 can be maximally 131 ( χ real ≤ Χ df ,90% ), the 2

2

global δ can be estimated according to:

(3)

δ≥

1 Χ

n

∑e

2 df , 90% i =1

2 i

The sum of squared residuals as derived from the data shown in Table S6 and S7 is 0.014. With Χ2110, 90 % = 131 the value of δ is estimated to be >= 0.010.

16

Schwender et al.: Supplemental Text

6. Model Validation One important prerequisite for the trustworthiness of the flux estimates is that quasi-steady state can be assumed. In addition, the model network can be validated by biochemical data (e.g. subcellular localization of enzymes) and by the labeling data itself. In the following sections, different assumptions that determine the network topology are described and justified. 6.1 Isotopic steady state. Isotopomer balancing under steady state relies on the isotopomer composition of metabolites being constant over time. During culture of the embryos the biomass doubled 4 to 5 times. Therefore at least 95 % of the biomass was formed during culture. The free amino acids are assumed to have a turnover time on the time scale of hours to days. Degradation of proteins and oil was assumed not to contribute significantly to the labeling of central metabolites. The isotopic steady state in free metabolites was assumed to be established after about 3 d of culture while the label in protein-bound amino acids after 14 d should average the label in free amino acids over the entire period of biomass increase. To test isotopic steady state over the labeling period the labeling in protein was compared to the label in free amino acids after 3 d. In the experiment with [U-13C6]Gln the label in protein bound amino acids after 14 d of growth was compared with the label in free amino acids extracted after a 3 d incubation (data not shown). 24 MS signals from 6 fragments were compared (18 degrees of freedom). With an allowed sum of squared differences of Χ2 18, 90 % = 26 the SD had to be 1.6% in the scale of fractional enrichments. Thus with the estimated SD in the system being the same number (see section 5.2) the labeling pattern observed after 3d in free metabolites and after 14 d in protein represented the same labeling pattern. Similar results were obtained with the labeling experiment using [1,2-13C2]/[U-13C6]glucose. The label accumulated in protein and the label in free amino acids after 14d were compared and were also consistent with one another on the basis of a 1.5 % SD. 6.2 Metabolic steady state. In order to maintain a constant concentration of each nutrient in the medium, embryos were cultured in a surplus of culture medium. In the experiments reported we observed biomass increase of about 0.5 mg DW per ml culture medium and according to uptake and biosynthesis rates shown in Fig. 1 (main text) none of the organic substrates in the medium was depleted by more than 10 %. In addition the uptake rates of the different carbon sources remained constant over time. This was concluded from the mass isotopomer distribution in molecular ions of labeled fatty acids according to (Schwender and Ohlrogge, 2002). After feeding [U-13C3]Ala for 14d, the mass isotopomer distribution observed in the molecular ion of C18:0 was in good agreement with the expected pattern due to the condensation of 9 acetate units being uniformly labeled to 7.5 % 13C (data not shown). Significant change in the uptake of [U13 C3]Ala relative to other carbon sources during growth would have resulted in subpopulations of fatty acids labeled differently, with the mass isotopomers in the molecular ion deviating from the expected binominal distribution of a polymer [acetate]9. Similarly, by feeding 13C-labeled glucose, carbon from glucose and sucrose will be mixed at the level of hexose phosphates. As shown before (Schwender and Ohlrogge, 2002) under the culture conditions used here over 90 % of the biomass is formed from a constant mixture

17

Schwender et al.: Supplemental Text of sucrose and glucose, resulting in constant dilution of the labeled glucose by hexose units derived from sucrose. We conclude that during growth the embryos take up all the different carbon sources in a constant ratio, allowing steady state flux analysis. We also noted that, if the medium concentrations of the organic nutrients are considered, Glc was preferentially taken up compared to Suc and Ala uptake was favored over Gln. 6.3 Isotopic equilibrium of transaminases. As described in section 4.1, based on 15NAla labeling the exchange flux between Alacyt and pyruvatecyt was found to be about 30 times the net flux. In a similar way the 13C-model assigned exchange between Gluc and KGm to be more than 10 times the net flux through these pools, indicating isotopic equilibration between the two. For Glu/KG this was independently confirmed from the 15 N-labeling experiments. Feeding 15N-Ala (together with 14N2-Gln) resulted in a 15Nenrichment of 45 % in the amino-N positions of all protein-bound amino acids, including Pro (data not shown). This isotopic equilibration between 15N and 14N in Pro shows that transaminase activities rapidly interconvert Glu with ketoglutarate. 6.4 Isotopic equilibrium among C4 dicarboxylic acids. The label in protein bound Asp was used in this study to reflect C4-dicarboxylic acids. Cytosolic Asp incorporated into protein is assumed to be derived from cytosolic OAA provided that abundant Asp transaminase activity is present in the cytosol. However, in Asp the labeling signature of symmetric randomization due to interconversion with fumarate was found. If embryos were labeled with [1,2-13C2]glucose / [U-13C6]glucose, central metabolism should generate the isotopomer [1,2,3-13C3]PEP in abundance which, after carboxylation by PEP carboxylase, should produce [1,2,3-13C3]OAA. Without additional reactions, [2,3,413 C3]OAA is not to be expected. However, the mass isotopomer distribution found in the fragments Asp(1-4), Asp(1-2) and Asp(2-4) shows that there was an almost equal abundance of the isotopomers [1,2,3-13C3] OAAcyt and [2,3,4-13C3]OAAcyt. This symmetric randomization of label in OAAc can be explained by the symmetry of fumarate if OAAc and fumarate are rapidly inter-converted. Since in plants fumarase has been found to be confined to mitochondria (Behal & Oliver, 1997), it can be concluded that cytosolic OAA isotopically equilibrates with mitochondrial fumarate by the combination of highly active dicarboxylic acid transporters, malate dehydrogenase and mitochondrial fumarase. Therefore, in the flux model mitochondrial and cytosolic OAA and malate were combined into one metabolic pool and the mitochondrial inter-conversion with fumarate was implemented as a free (a priori unknown) exchange flux. The resulting exchange flux was the biggest flux in the system after vAATXCH. Randomization of label in OAA/malate by fumarase has also been observed in maize root tips (Salon et al., 1988; Dieuaide-Noubhani et al., 1995; Edwards et al., 1998), in tomato cell culture (Rontein et al., 2002) and in animal systems (Des Rosiers et al., 1994). 6.5 Isotopic equilibrium of cytosolic and plastidic phosphoenol pyruvate. The only metabolites representing PEP that are accessible in the MS data are the aromatic amino acids. However, these are formed in the plastid and therefore represent plastidic PEP. No labeling information on cytosolic PEP is present and the isotopic equilibration of plastidic and cytosolic pools cannot be directly determined. However, according to Kubis et al. (2004), in isolated plastids of developing B. napus embryos at mid-oil stage the capacity

18

Schwender et al.: Supplemental Text of PEP import matches the in vivo rate of fatty acid synthesis. This suggests that the PEP translocator (PPT) is abundant. Therefore we assume isotopic equilibration of PEPcyt and PEPpl by bidirectional exchange between cytoplasm and plastid, and we unify both into one PEP pool in the flux model. This assumption is also supported from the way that PEP is produced. PEP is formed from PGA either in the cytosol or in the plastid (formation of PEP from pyruvate or OAA could be excluded based on labeling experiments, see section 1.1). PGA itself can be rapidly exchanged between cytosol and plastid by the reversible triose phosphate translocator (TPT) which is also abundant in developing seeds of B. napus (Gupta and Singh, 1996) and is expressed in abundance in Arabidopsis developing seeds (Ruuska et al., 2002). Thus both TPT and PPT will contribute to the isotopic equilibration of the cytosolic and plastidic PEP pools. 6.6 The pyruvate carrier. Uptake kinetics with isolated plastids suggest the presence of a plastidic pyruvate carrier in B. napus embryos (Eastmond and Rawsthorne, 2000). As has been shown in the case of the mitochondrial pyruvate carrier (Laloi, 1999), both the mitochondrial and the plastidic pyruvate carrier were assumed to be driven by pH gradients and therefore to work unidirectionally, having essentially no exchange flux. Furthermore, the observed difference in label between (cytosolic) Ala and Val (derived from plastid pyruvate) demonstrates that cytosolic and plastidic pyruvate are not isotopically equilibrated. Therefore, in the flux model the import of pyruvate into the plastids and mitochondria was modeled as non-reversible. 6.7 Independent confirmation of the uptake ratio of Ala and Gln by 15N labeling. Ala and Gln were the sole nitrogen sources in the culture medium. After labeling B. napus embryos with (amide) 15N-Gln, (amino) 15N-Gln and 15N-Ala (all 99 % 15N enriched), the 15 N enrichment in proteinogenic amino acids was determined by GC/MS. After labeling with 15N-Ala a 15N-enrichment of 45 % was consistently found in the amino-N positions of all amino acids (Data not shown). The results from (amide)-15N-Gln and (amino)-15NGln labeling experiments show that 45 % of the protein bound nitrogen positions are derived from Ala and 55 % derived from Gln. Taking into account that Ala and Gln contain one and two nitrogens respectively, the labeling results can be explained by an uptake ratio of Ala:Gln being 62:38 (mol/mol). This ratio was confirmed independently by the 13C-studies, where the influx of Ala and Gln were determined to be in the ratio 63:37 (Fig. 1, main text). 6.8 Influence of atmospheric CO2. Although there is a net output of CO2 from the embryo cultures, the influence of unlabeled CO2 from the atmosphere on the results of labeling experiments has to be considered. We have performed labeling experiments under identical conditions as reported in the paper except that to an unlabeled growth medium 13C-enriched CO2 (99% 13C enrichment) was given at a 2 % (v/v) concentration in the air space above the medium. We found in all analyzed metabolites a maximum positional enrichment of 12 % 13C. Considering Fick’s 1st law of diffusion and a constant CO2 concentration in the embryos, the diffusion rate of atmospheric CO2 into the embryos will change in linear correlation to it’s concentration in the air. In this case the diffusion rate of CO2 at ambient 0.037 % in air (370 ppm) would be 57 times less than the diffusion rate of CO2 concentrated at 2 % (2% / 0.037%). Considering the maximal

19

Schwender et al.: Supplemental Text 12% positional 13C-enrichment in metabolites under 2 % 13CO2 in air, this means the fraction of 12C in any C-position in the network derived by air-borne 12CO2 would be increased by maximal 0.2 % in the fractional enrichment scale (12% 13C / 57 = 0.2% 13C, or 12C in the case of unlabeled ambient CO2). Therefore the influence of atmospheric CO2 on the labeling pattern can be assumed to be insignificant and ignored in isotopomer analysis.

20

Schwender et al.: Supplemental Text

7. Supplemental Tables

TAG Total protein Cell wall / starch other compounds

% (w/w) of biomass 38 % 17 % 37 % 8%

SD 3.8 % 1.7 % 3.7 %

Table S1 Biomass composition of cultured B. napus embryos. Main storage compounds in embryos grown under conditions identical to this study. In short, embryos were homogenized and extracted in hexane/isopropanol (2:1) (Lipid extract) and methanol/water (8:2). After transesterification of the lipid fraction, lipid was quantified by gas chromatography of fatty acid methyl esters with glyceryl triheptadecanoate as internal standard. In the remaining cell pellet protein was determined after extraction (1 % sodium dodecyl sulfate, 1 mM β-mercaptoethanol, 50 mM Tris pH 8.0) by BCA assay (Sigma, St. Louis) and confirmed by C/N analysis of dry embryos. The weight of the cell pellet minus protein was assumed to be a glucose polymer with molecular weight of 162 g/mol of the monomers.

(% w/w) Ala Arg Asn Asp Cys Glu Gln Gly His Ile Leu Lys Meth Phe Pro Ser Thr Trp Tyr Val

Main seed protein components Cruciferin Napin Oleosin 60 ± 6 20 ± 2 20 ± 2 6.7 5.7 5.9 4.0 1.0 3.8 12.3 10.7 2.0 4.8 8.3 3.0 1.2 4.0 5.0 7.7 3.8 1.0 2.2 6.9

6.2 3.9 3.9 3.9 4.5 5.1 14.6 5.6 1.7 2.8 7.9 5.6 2.8 5.6 7.9 5.1 5.1 1.1 1.7 5.1

7.4 8.6 0.6 8.6 0.0 1.1 6.3 10.3 1.1 6.3 8.6 4.6 2.9 2.3 2.9 7.4 8.6 0.6 5.7 6.3

Composition of seed protein (mol %) 6.7 ± 0.1 5.7 ± 0.2 4.6 ± 0.3 4.7 ± 0.3 1.6 ± 0.1 3.7 ± 0.2 11.8 ± 0.4 9.6 ± 0.1 1.8 ± 0.1 4.6 ± 0.1 8.2 ± 0.1 3.8 ± 0.1 1.8 ± 0.1 4.1 ± 0.1 5.3 ± 0.3 7.1 ± 0.1 4.8 ± 0.3 1.0 ± 0.1 2.6 ± 0.2 6.4 ± 0.1

Table S2 Amino acid composition of proteins in B. napus embryos. The seed storage protein in B. napus consists of 60 % cruciferin, 20 % napin and 20 % oleosin (Norton, 1989). The amino acid compositions of the three proteins were derived from sequence data (NCBI protein database). The amino acid composition of seed protein was calculated considering the amino acid compositions (mol %) of the three protein components and assuming a 10 % SD in the values for the protein fractions. The resulting amino acid composition is similar to the analysis of total seed protein reported by Norton (1989).

21

Schwender et al.: Supplemental Text

Biosynthetic precursor demands

Fatty acid compostion Fatty acid C16:0 C18:0 C18:1 C18:2 C18:3 C20:0 C20:1 C22:0 C22:1

Mol % 7.4 ± 1.4 1.9 ± 0.3 16.2 ± 2.8 19.6 ± 2.7 15.3 ± 1.5 1.0 ± 0.2 9.1 ± 0.8 1.0 ± 0.2 28.6 ± 1.6

glycerol

100

Acetatepl 24 27 27 27 27 27 27 27 27

Acetatecyt 0 0 0 0 0 3 3 6 6

glycerol

1

Total precursor demand per mol TAG

26.8 ±1.2

2.1 ±0.2

1

Table S3 Fatty acid composition of TAG and biosynthetic precursor demands. The fatty acid composition (Mol %) of cultured B. napus embryos (cv. Reston) was analyzed by gas chromatography (Data taken from Schwender et al., 2002). Also biosynthetic demands of plastidic and cytosolic acetyl-CoA are given: For each fatty acid species the amount of cytosolic or plastidic acetyl-CoA needed for the synthesis of one molecule triacylglycerol is given. C20 and C22 are derived by cytosolic elongation (cytosolic acetyl-CoA) of C18 fatty acids by one and two units of cytosolic acetyl-CoA, respectively. The SDs of precursor demands are derived from the SD in the fatty acids composition by a Monte Carlo approach.

22

Schwender et al.: Supplemental Text

GC/MS Fragment Metabolic precursor (part of the precursor that measured contributes) Ala(1-3) Pyrcyt (1-3) Ala(2-3) Pyrcyt (2-3) C18(1-2) AcCoApl (1-2) AcCoAcyt (1-2) C22(1-2) Pyrpl (1-3) x Pyrpl (2-3) Val(1-5) Val(2-5) Pyrpl (2-3) x Pyrpl (2-3) Glucyt (1-5) Pro(1-5) Glucyt (2-5) Pro(2-5) Asp(1-4) OAAcyt (1-4) OAAcyt (1-2) Asp(1-2) OAAcyt (2-4) Asp(2-4) Table S4 Precursor – product relationship according to biosynthetic relations and carbon transitions. Suffixes denote carbon atoms which are referred to. Suffix “cyt” = cytosolic; “pl” = plastidic. See also Fig. S2.

Monomer

precursor from central metabolism AcCoApl

AcCoAcyt

Pyrpl

Pyrcl

PEP

OAAcyt

Glucyt

Glncyt

CO2

E4P

PGA

DHAP

PentP

Ala -1 Arg -1 -1 Asn -1 Asp -1 Cys -1 Glu -1 Gln -1 Gly 1 -1 His -1 Ile -1 -1 1 Leu -1 -2 1 Lys -1 -1 1 Meth -1 -1 Phe -2 1 -1 Pro -1 Ser -1 Thr Trp 1 -2 1 -1 -1 -1 Tyr -2 1 -1 Val -2 1 -1 acetatepl -1 acetatecyt glycerol -1 Table S5 Biosynthetic stoichiometric relations between metabolites of central metabolism and monomers of storage polymers (Protein, TAG). This table is a matrix used to translate the fractions of monomers in biomass (Tables S1, S2) into the biosynthetic demands of intermediate precursors. Information on the pathways can be seen at http://www.arabidopsis.org:1555/ARA/

23

Schwender et al.: Supplemental Text Labeled precursors

Fragment Ala(1-3)

Ala(2-3)

C18(1-2)

C22(1-2)

Val(1-5)

Val(2-5)

Pro(1-5)

Pro(2-5)

Asp(1-4)

Asp(1-2)

Asp(2-4)

m0 m1 m2 m3 m0 m1 m2 m0 m1 m2 m0 m1 m2 m0 m1 m2 m3 m4 m5 m0 m1 m2 m3 m4 m0 m1 m2 m3 m4 m5 m0 m1 m2 m3 m4 m0 m1 m2 m3 m4 m0 m1 m2 m0 m1 m2 m3

[1,2-13C2]glucose/ [U-13C6]glucose

[U-13C3]Ala

[U-13C5]Gln, 50 %

Measured

Predicted

Measured

Predicted

Measured

Predicted

0.814 0.046 0.069 0.071 0.850 0.011 0.139 0.761 0.032 0.207 0.879 0.027 0.094 0.541 0.065 0.236 0.101 0.028 0.029 0.564 0.046 0.328 0.007 0.055 0.918 0.022 0.044 0.011 0.004 0.002 0.930 0.014 0.051 0.001 0.004 0.705 0.135 0.083 0.065 0.011 0.756 0.098 0.106 0.040 0.780 0.160 0.060

0.801 0.042 0.078 0.079 0.828 0.019 0.153 0.770 0.025 0.205 0.863 0.023 0.115 0.565 0.061 0.233 0.095 0.024 0.022 0.592 0.038 0.317 0.010 0.042 0.890 0.030 0.059 0.014 0.005 0.002 0.906 0.018 0.068 0.003 0.006 0.695 0.141 0.089 0.064 0.011 0.758 0.096 0.107 0.039 0.778 0.163 0.059

0.675 0.017 0.020 0.289 0.687 0.016 0.297 0.922 0.006 0.072 0.878 0.010 0.112 0.834 0.018 0.073 0.069 -0.001 0.008 0.842 0.010 0.140 -0.001 0.007 0.914 0.015 0.065 0.004 0.000 0.002 0.927 0.007 0.066 0.000 0.000 0.893 0.090 0.012 0.004 0.000 0.936 0.056 0.007 0.001 0.949 0.043 0.009

0.678 0.013 0.009 0.300 0.690 0.006 0.303 0.924 0.002 0.074 0.857 0.009 0.134 0.838 0.018 0.069 0.069 0.000 0.005 0.853 0.004 0.137 0.000 0.006 0.930 0.017 0.049 0.003 0.000 0.000 0.942 0.008 0.049 0.001 0.000 0.889 0.092 0.018 0.000 0.000 0.933 0.059 0.009 0.000 0.945 0.047 0.008

0.991 0.004 0.002 0.003 0.993 0.002 0.005 0.989 0.004 0.006 0.844 0.017 0.139 0.982 0.005 0.009 0.003 0.000 0.000 0.981 0.007 0.012 0.000 0.000 0.590 0.013 0.018 0.020 0.021 0.337 0.602 0.007 0.045 0.019 0.326 0.804 0.029 0.029 0.099 0.041 0.807 0.021 0.070 0.102 0.821 0.089 0.090

0.998 0.002 0.000 0.000 1.000 0.000 0.000 1.000 0.000 0.000 0.833 0.010 0.156 0.997 0.003 0.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 0.000 0.574 0.008 0.032 0.029 0.018 0.338 0.576 0.010 0.058 0.015 0.342 0.803 0.018 0.030 0.103 0.045 0.810 0.022 0.062 0.106 0.823 0.081 0.096

Table S6 GC/MS measurements used for flux analysis. Measured and model predicted labeling in amino acids and fatty acids, labeled in three labeling experiments with differently labeled precursors. Predicted values correspond to the best flux estimates. MS measurements are corrected for natural abundance of 13C and for mass isotope contribution of hetero-atoms and of the derivative side chains. Values are normalized (sum of all peaks of one fragment = 1).

24

Schwender et al.: Supplemental Text Continuation of Table S6 The instrument related SD for MS measurements average 0.0015 +/- 0.003 while the error in between the three experiments is higher by about an order of magnitude. Boldface numbers: Largest fitting errors for these measurements. These 13 measurements constitute 50 % of the Χ2 sum for of errors.

fragment m1RP(1-5) m0 m1 m2 m3 m4 m5 m0 m1TP(1-3) m1 m2 m3 m1HP(1-6) m0 m1 m2 m3 m4 m5 m6 m1Gln(1-5) m0 m1 m2 m3 m4 m5

measured

predicted

0.551 0.575 0.095 0.073 0.188 0.189 0.087 0.084 0.042 0.043 0.034 0.039 0.710 0.725 0.050 0.042 0.112 0.107 0.128 0.126 0.578 0.588 0.043 0.039 0.169 0.175 0.073 0.076 0.032 0.029 0.021 0.017 0.083 0.076 0.975 0.975 0.004 0.007 0.014 0.014 0.004 0.003 0.001 0.001 0.002 0.000 Table S7 Measurements considered for the label experiment [1,2-13C2]glucose/ [U-13C6]glucose in addition to the data shown in table S6. The fragment m1Gln(1-5) is derived from a measurement of free Gln.

Metabolites analyzed

Labeled Precursors N-Ala* [1-15N]-Gln [5-15N]-Gln Ala 44.2 ± 0.4 % 31.9 % 17.4 % Val, Phe, Gly, Ser, Pro 42.1 ± 0.5 % 33.5± 0.2 % 18.1± 0.1 % Table S7a Average 15N enrichment in different amino acids after feeding differently 15N-labeled Ala and Gln under the same conditions as in the 13C-labeling experiments. Ala or Gln in the medium were enriched to 99 % 15N in the 1– or 5-position. *n = 3 experiments. 15

25

Schwender et al.: Supplemental Text

Flux name

NET flux

vuptGlc vG6PDH vRub Valdo vGAPDH vPGM vuptAla vuptGln vPKc vPKp vAAT vGlnGlu vPEPC vME vCS vACL vICDH vKDH vFM vGAT vPyr1 vPyr2 vPDH vCO2_out vFASt vFASpl vFASc vAla_P vPyrpl_P vPEP_P vAcCoApl_P vAsp_P vGlu_P vGln_P vHis_P vEP_P vSer_P vGlyc_TAG vHP_St

58.3 5.6 19.3 31.6 46.2 83.0 7.4 4.3 17.9 60.0 6.6 3.0 3.6 2.3 5.8 5.9 -0.1 1.3 1.3 1.4 21.0 3.5 77.0 71.0 82.0 76.1 5.9 0.7 4.0 1.6 0.9 2.6 1.6 1.3 0.3 0.8 2.0 2.8 15.9

SD

4.2 (4.5*) 2.3 (3.3) 3.5 (4.7) 3.6 (3.9) 7.9 (10.6) 8.3 (8.7) 0.7 (2.1) 0.3 (0.5) 2.0 (5.1) 6.5 (9.3) 0.7 (2.1) 0.2 (0.4) 0.3 (1.0) 0.3 (1.3) 0.6 (0.7) 0.6 (0.7) 0.2 (0.4) 0.2 (0.5) 0.2 (0.5) 0.2 (0.4) 2.6 (7.0) 0.3 (1.3) 8.9 (9.1) 8.7 (10.6) 8.9 (9.0) 8.9 (9.1) 0.6 (0.7) 0.1 (0.1) 0.6 (0.6) 0.2 (0.2) 0.1 (0.1) 0.3 (0.4) 0.2 (0.2) 0.2 (0.2) 0.0 (0.0) 0.1 (0.1) 0.3 (0.3) 0.3 (0.3) 1.4 (1.4)

Exchange flux

68 % conficence interval lower upper bound bound

43.8 65.4

35.8 6.5

52.4 185.8

190.0 1.3

106 0.7

274 1.8

4.5

3.5

5.4

29.5 4.7

13.6 3.6

46.9 5.8

Table S8 Best Fit flux estimates. This set of flux values [nmol h-1 mg FW-1] was obtained consistently by starting the flux parameter fitting with random start values over a hundred times. SD in fluxes are given for a SD of 1.0 % in the MS measurements. For exchange fluxes upper and lower bounds of the unsymmetrical 68 % confidence intervals are given. * In parentheses are the SD values obtained if only the MS data from the labeling with [1,2-13C2]glucose/ [U13 C6]glucose are considered.

26

Schwender et al.: Supplemental Text

Flux name

Reaction

Enzyme

Net flux

NADH

FADH

ATP

Mitochondrial ATP production by substrate oxidation vME vCS vPDHmit vICDH vKGDH vFM

malate Æ Pyrmit + NADH OAA + AcCoAmit Æ Cit Pyrmit Æ AcCoAmit + CO2 + NADH Cit Æ KG + CO2 + NADH KG Æ Succ + CO2 + NADH + GTP Succ Æ Fum + FADH Fum Æ mal Mal Æ OAA + NADH

Mitochondrial malic enzyme Citrate synthase Plastidic pyruvate dehydrogenase complex Aconitase, Isocitrate dehydrogenase Ketoglutarate dehydrogenase Succinyl-CoA synthethase Succinate dehydrogenase fumarase malate dehydrogenase

2..3 5.8 5.8

2.3 5.8

7.0 0.0 17.5

-0.1

-0.1

-0.2

1.3

1.3

4.0

1.3 1.3

1.3 1.3

ATP produced in mitochondria by oxidative phosphorylation from mitochondrial NADH and FADH*1

2.6 4.0 34.8

ATP need for Fatty acid and Protein Synthesis vFASpl

AcCoApl Æ FA

vFASc

AcCoAcyt + FA Æ VLCFA

vACL

Cit Æ AcCoAcyt +OAA

Plastidic fatty acid synthesis to C16 and C18 fatty acids Cytosolic elongation of C18 fatty acids to C20 and C22 Cytosolic ATP:Citrate lyase

76.1

76.1

5.9

5.9

5.9

5.9

ATP needed for fatty acid synthesis amino acids into protein*2

16.6

87.9

X 4.3 mole ATP / mol AA*3

71.3

Total ATP need for protein and TAG synthesis

159.2

Table S9. Estimation of ATP production by mitochondria and ATP demand by biosynthesis of protein and TAG. Flux values (nmol h-1 mgFW-1) for B. napus embryos taken from table S8. Cofactor balances are given according to fluxes. *1 maximal possible ATP/O ratios were assumed as 3.0 for NADH and 2.0 for FADH *2 sum of fluxes into biomass for all amino acids, see Table S2. *3 estimated cost for protein synthesis was assumed as 4.3 moles ATP per mole amino acid incorporated (Stefanopoulos et al., 1998)

27

Schwender et al.: Supplemental Text Table S10 Network definition, extract from the model file used in 13CFLUX. The first subnetwork is shown, simulating labeled glucose. NETWORK FLUX_NAME

EDUCT_1

EDUCT_2

PRODUCT_1

PRODUCT_2

// Uptake of substrates vuptU

vupt0

vupt

vuptAla

vuptGln

m1GLCU

m1GLC

#ABCDEF

#ABCDEF

//uptake labeled glucose

m1GLC0

m1GLC

#ABCDEF

#ABCDEF

m1GLC

m1HP

#ABCDEF

#ABCDEF

m1Ala_in

m1AlaCyt

#ABC

#ABC

m1Gln_in

m1Gln

#ABCDE

#ABCDE

m1HP

m1CO2

m1RuP

#ABCDEF

#A

#BCDEF

//uptake unlabeled glucose

//uptake alanine

//uptake glutamine

// calvin cycle / ppp vG6PDH

vTK1

vSaldo

vTK2

vPPiso

vPPepi

vRub

m1XP

m1RP

m1SP

m1TP

#ABCDE

#abcde

#ABabcde

#CDE

m1SP

m1EP

m1TP

// SHP2 aldolase, reversible

#ABCDEFG

#DEFG

#CBA

// with Sh bisphosphatase // transketolase

m1XP

m1EP

m1TP

m1HP

#ABCDE

#abcd

#CDE

#ABabcd

m1RuP

m1RP

#ABCDE

#ABCDE

m1RuP

m1XP

#ABCDE

#ABCDE

// Ru5P epimerase

// Ru5P isomerase

m1RuP

m1CO2

m1PGA

m1PGA

#ABCDE

#a

#aBA

#CDE

28

// transketolase

// RubisCO

Schwender et al.: Supplemental Text Continuation of table S10 // Embden Meyerhof Pathway valdo

m1HP #ABCDEF

m1TP #CBA

vGAPDH

m1TP #ABC

m1PGA #ABC

// GAP-dehydrogenase

vPGM

m1PGA #ABC

m1PEP #ABC

// P-glycerate mutase, enolase

m1PyrCyt #ABC

// Ala / XXX transaminase

m1PyrCyt #ABC

// cytosolic pyruvate kinase // compartmentation of PEP is not considered

m1PyrPl #ABC

// plastidic pyruvate kinase

m1OAA #ABCa

// PEP carboxylase

vAAT

m1AlaCyt #ABC // cytosolic and plastidic metabolism of PEP, Pyruvate, Ala, Glu vPKc m1PEP #ABC

m1TP #DEF

// F1,6P2 aldolase

vPKpl

m1PEP #ABC

vPEPC

m1PEP #ABC

vGlnGlu

m1Gln #ABCDE

m1Glu #ABCDE

// Glutaminase // glutamate synthase

vGAT

m1Glu #ABCDE m1PyrCyt #ABC

m1KG #ABCDE m1PyrPl #ABC

// glutamate/KG transaminase // and exchange cytosol / mitochondrium // pyruvate import into plastid

vPyr2

m1PyrCyt #ABC

m1Pyrm #ABC

// Import of pyruvate into mitochondria

vVal

m1PyrPl #ABC

vPDHpl

vPyr1

m1CO2 #a

m1PyrPl #abc

m1Val #ABbcC

m1CO2 #a

// plastidic valine synthesis

m1PyrPl #abc

m1AcPl #bc

m1CO2 #a

// formation of plastidic acetyl-CoA // major CO2 producer

vFASPl

m1AcPl #AB

m1AcFA #AB

// plastidic fatty acid synthesis // AcFA is also fueled from AcCyt

vAco

m1Cit #ABCDEF

m1Icit #EDCBAF

// Aconitate hydratase

29

Schwender et al.: Supplemental Text Continuation of table S10 vACL

m1Cit #ABCDEF

m1OAA #FCDE

m1AcCyt #AB

// ATP:citrate lyase

vICDH

m1Icit #ABCDEF

m1KG #ABCDE

m1CO2 #F

// isocitrate dehydrogenase

vKDHa

m1KG #ABCDE m1KG #ABCDE

m1Succ #BCDE m1Succ #EDCB

m1CO2 #A m1CO2 #A

// ketoglutarate dehydrogenase

vFM1

m1Succ #ABCD

m1OAA #ABCD

// Fumarase, malate dehydrogenase

vFM2

m1Succ #ABCD

m1OAA #DCBA

// two reactions for symmetric // randomization in succinate

vME

m1OAA #ABCD m1AcCyt #AB

m1Pyrm #ABC m1AcFA #AB

vCO2_out

m1CO2 #A

m1CO2_out #A

// CO2 release

vAla_P

m1AlaCyt #ABC

m1AlaP #ABC

// Ala into protein

vPyrpl_P

m1Val #ABCDE

m1ValP #ABCDE

// Val into protein // also Leu, Ile, Lys

vAsp_P

m1OAA #ABCD

m1Asp #ABCD

// Asp into protein // also Asn, Thr, Met, Ile, Lys

vGlu_P

m1Glu #ABCDE

m1Gluprot #ABCDE

// Glu into protein // also Pro, Arg

vGln_P

m1Gln #ABCDE

m1Glnprot #ABCDE

// Gln into protein

vFAS

m1AcFA #AB

m1AcFA_out #AB

// Acetate units into fatty acids // plastidic FA synthesis + cytosolic elongation

vPEP_P

m1PEP #ABC

m1PEP_P #ABC

// PEP into Phe, Tyr, Trp (protein)

vKDHb

vFAScyt

m1CO2 #D

// two reactions for symmetric // randomization in succinate

// mitochondrial malic enzyme // cytosolic fatty acid elongation // AcFA is also fueled from AcPl

// output

30

Schwender et al.: Supplemental Text Continuation of table S10 vAcpl_P

m1AcPl #AB

m1Ac_P #AB

// Acpl into Leu (protein)

vHis_P

m1RP #ABCDE

m1PrPP #ABCDE

// RP into His, Trp

vEP_P

m1EP #ABCD m1PGA #ABC

m1Aro #ABCD m1Ser #ABC

// E4P into aromatic AA

vGlyc_TAG

m1TP #ABC

m1glycerol #ABC

// glycerol part of TAG

vHP_St

m1HP #ABCDEF

m1St #ABCDEF

// HP into starch and cell wall polymers

vSer_P

31

// PGA into Ser, Gly, Cys

Schwender et al.: Supplemental Text

8. References Box, G.E.P., and Müller, M.E. (1958) A note on the generation of random normal deviates. Annals Math. Stat 29, 610-611 Behal, R.H., and Oliver, D.J. (1997) Biochemical and molecular characterization of fumarase from plants: Purification and characterization of the enzyme—cloning, sequencing, and expression of the gene. Arch Biochem Biophys 348, 65–74 Chia, T.Y.P., Pike, M.J., and Rawsthorne, S. (2005) Storage oil breakdown during embryo development of Brassica napus (L.). J. Exp. Bot. 56, 1285-1296 Dennis, D. T. (1989) in Physiology, Biochemistry, and Genetics of Nongreen Plastids (Boyer, C. D., Shannon, R. C., and Hardison, R. C., eds) pp. 120–129, American Society of Plant Physiologists, Rockville, MD Des Rosiers, C., Fernandez, C.A., David, F., and Brunengraber, H. (1994) Reversibility of the mitochondrial isocitrate dehydrogenase reaction in the perfused rat liver. Evidence from isotopomer analysis of citric acid cycle intermediates. J. Biol. Chem. 269, 27179-27182 Dieuaide-Noubhani, M., Raffard, G., Canioni, P., Pradet, A., and Raymond, P. (1995) Quantification of compartmented metabolic fluxes in maize root- tips using isotope distribution from C-13-labeled or C-14- labeled glucose. J. Biol. Chem. 270, 1314713159 Domergue, F., Cassagne, C., and Lessire, R. (1999) Seed acyl-coa elongases: the other system of fatty acid synthesis. Ocl-Ol. Corps Gras Lipides 6, 101–106 Eastmond, P. J., and Rawsthorne, S. (2000)Coordinate changes in carbon partitioning and plastidial metabolism during the development of oilseed rape embryo. Plant Physiol. 122, 767-774 Edwards, S., Nguyen, B. T., Do, B., and Roberts, J. K. M. (1998) Contribution of malic enzyme, pyruvate kinase, phosphoenolpyruvate carboxylase, and the Krebs cycle to respiration and biosynthesis and to intracellular ph regulation during hypoxia in maize root tips observed by nuclear magnetic resonance imaging and gas chromatography-mass spectrometry. Plant Physiol. 116, 1073-1081 Goffman, F. D., Ruckle, M., Ohlrogge, J. B., and Shachar-Hill, Y. (2004) Carbon dioxide concentrations are very high in developing oil seeds. Plant Physiol. Biochem. 42, 703-708

Goffman, F.D., Alonso, A.P., Schwender, J., Shachar-Hill, Y., and Ohlrogge, J.B. (2005) Light enables a very high efficiency of carbon storage in developing embryos of rapeseed. Plant Physiol. 138, 2269-2279 Gupta, R., and Singh, R. (1996) Fatty acid synthesis by isolated leucoplasts from developing Brassica seeds: role of nucleoside triphosphates and DHAP- shuttle as the source of energy. J. Biosciences 21, 819-826 Ireland, R. J., and Lea, J. P. (1999) in Plant Amino Acids: Biochemistry and Biotechnology (Singh, B. K., ed.) pp. 49–109, Marcel Dekker, New York Kubis, S.E., Pike, M.J., Everett, C.J., Hill, L.M., and Rawsthorne, S. (2004) The import of phosphoenolpyruvate by plastids from developing embryos of oilseed rape, Brassica 32

Schwender et al.: Supplemental Text napus (L.), and its potential as a substrate for fatty acid synthesis. J. Exp. Bot. 55, 1455– 1462 Ho, C. L., Noij, M., and Saito, K. (1999) Plastidic pathway of serine biosynthesis. J. Biol. Chem. 274, 11007–11012 Laloi, M. (1999) Plant mitochondrial carriers: an overview. Cell. Mol. Life Sci. 56, 918–944 Mollney, M., Wiechert, W., Kownatzki, D., and de Graaf, A.A. (1999) Bidirectional reaction steps in metabolic networks part IV: Optimal design of isotopomer labeling experiments. Biotechnol. Bioeng. 66, 86–103 Norton, G. Nature and biosynthesis of storage proteins. In G Röbbelen, RK Downey, A Ashri, eds., Oil Crops of the World. McGraw Hill, NY, pp 165-191 (1989) Ohta, D., Fujimori, K., Mizutani, M., Nakayama, Y., Kunpaisal-Hashimoto, R., Münzer, S., and Kozaki, A. (2000) Molecular cloning and characterization of ATP-phosphoribosyl transferase from Arabidopsis, a key enzyme in the histidine biosynthetic pathway. Plant Physiol. 122, 907–914 Ruuska, S., Girke, T., Benning, C., and Ohlrogge, J.B. (2002) Contrapuntal networks of gene expression during Arabidopsis seed filling. Plant Cell 14, 1191-1206 Ruuska, A.S., Schwender, J., and Ohlrogge, J.B. (2004) The capacity of green oilseeds to utilize photosynthesis to drive biosynthetic processes. Plant Physiol. 136, 2700-2709 Rontein, D., Dieuaide-Noubhani, M., Dufourc, E. J., Raymond, P., and Rolin, D. (2002) The metabolic architecture of plant cells. Stability of central carbon metabolism and flexibility of anabolic pathways during the growth cycle of tomato cells. J. Biol. Chem. 46, 43948-43960 Salon, C., Raymond, P., and Pradet, A. (1988) Quantification of carbon fluxes through the tricarboxylic acid cycle in early germinating lettuce embryos. J. Biol. Chem. 263, 1227812287 Schultz, C. J., and Coruzzi, G. M. (1995) The aspartate aminotransferase gene family of Arabidopsis encodes isoenzymes localized to three distinct subcellular compartments. Plant J. 7, 61–75 Schwender, J. and Ohlrogge, J.B. (2002) Probing in vivo metabolism by stable isotope labeling of storage lipids and proteins in developing Brassica napus embryos. Plant Physiol. 130, 347-361 Schwender, J., Ohlrogge, J.B. and Shachar-Hill, Y. (2003) A flux model of glycolysis and the oxidative pentosephosphate pathway in developing Brassica napus embryos. J. Biol. Chem. 278, 29442-29453 Schwender, J., Goffman, F., Ohlrogge, J.B. and Shachar-Hill, Y. (2004) RuBisCO without the Calvin cycle improves the carbon efficiency of developing green seeds. Nature 432, 779782 Schwender, J., Ohlrogge, J. and Shachar-Hill Y (2004b) Understanding flux in plant metabolic networks. Current Opinion Plant Biol. 7, 309-317

33

Schwender et al.: Supplemental Text Singh, B. K. (1999) in Plant Amino Acids: Biochemistry and Biotechnology (Singh, B. K., ed) pp. 227–247, Marcel Dekker, New York Stefanopoulos, G. N., Aristidou, A. A. and Nielsen, J. (1998) Metabolic Engineering: Principles and Methodologies, Academic Press, London Szyperski T (1998) C-13-NMR, MS and metabolic flux balancing in biotechnology research. Q. Rev. Biophys. 31, 41–106 Wiechert, W. and de Graaf, A. (1997) Bidirectional reaction steps in metabolic networks: I. Modeling and simulation of carbon isotope labeling experiments. Biotechnol. Bioeng. 55, 101–117 Wiechert, W., Mollney, M., Isermann, N., Wurzel, M. and de Graaf, A.A. (1999) Bidirectional reaction steps in metabolic networks: III. Explicit solution and analysis of isotopomer labeling systems. Biotechnol. Bioeng. 66, 69-85 Wiechert, W., Siefke, C., de Graaf, A.A. and Marx, A (1997) Metabolic Networks: II. Flux estimation and statistical analysis. Biotechnol. Bioeng. 55, 118–135 Wiechert, W., Mollney, M., Petersen, S. and de Graf, A.A. (2001) A universal framework for 13C metabolic flux analysis. Metabol. Eng. 3, 265-283 Wilkie, S. E. and Warren, M. J. (1998) Recombinant expression, purification, and characterization of three isoenzymes of aspartate aminotransferase from Arabidopsis thaliana. Protein Expression Purif. 12, 381–389 Whitfield, H. V., Murphy, D. J., and Hills, M. J. (1993) Subcellular-localization of fatty-acid elongase in developing seeds of Lunaria annua and Brassica napus. Phytochemistry 32, 255–258

34

Mitochondrial Metabolism in Developing Embryos of Brassica napus Jörg Schwender, Yair Shachar-Hill and John B. Ohlrogge J. Biol. Chem. 2006, 281:34040-34047. doi: 10.1074/jbc.M606266200 originally published online September 12, 2006

Access the most updated version of this article at doi: 10.1074/jbc.M606266200 Alerts: • When this article is cited • When a correction for this article is posted Click here to choose from all of JBC's e-mail alerts

http://www.jbc.org/content/suppl/2006/09/13/M606266200.DC1.html This article cites 44 references, 24 of which can be accessed free at http://www.jbc.org/content/281/45/34040.full.html#ref-list-1

Downloaded from http://www.jbc.org/ by guest on May 16, 2016

Supplemental material: